messages
listlengths 3
3
|
|---|
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/02b4262a083537a9.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Man</name>\n <bndbox>\n <xmin>0.29</xmin>\n <ymin>0.22</ymin>\n <xmax>0.71</xmax>\n <ymax>0.98</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.32</xmin>\n <ymin>0.45</ymin>\n <xmax>0.73</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/06a7a62c2e31dd4e.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Musical keyboard</name>\n <bndbox>\n <xmin>0.16</xmin>\n <ymin>0.58</ymin>\n <xmax>0.74</xmax>\n <ymax>0.75</ymax>\n </bndbox>\n </object>\n <object>\n <name>Microphone</name>\n <bndbox>\n <xmin>0.44</xmin>\n <ymin>0.28</ymin>\n <xmax>0.52</xmax>\n <ymax>0.33</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/027eec316d7e8a85.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Motorcycle</name>\n <bndbox>\n <xmin>0.43</xmin>\n <ymin>0.46</ymin>\n <xmax>0.57</xmax>\n <ymax>0.62</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/034ee3cc5ff5eaaf.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Suit</name>\n <bndbox>\n <xmin>0.79</xmin>\n <ymin>0.21</ymin>\n <xmax>0.94</xmax>\n <ymax>0.89</ymax>\n </bndbox>\n </object>\n <object>\n <name>Woman</name>\n <bndbox>\n <xmin>0.02</xmin>\n <ymin>0.19</ymin>\n <xmax>0.18</xmax>\n <ymax>0.95</ymax>\n </bndbox>\n </object>\n <object>\n <name>Woman</name>\n <bndbox>\n <xmin>0.83</xmin>\n <ymin>0.18</ymin>\n <xmax>0.99</xmax>\n <ymax>0.92</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/063812ed1cfcb7ed.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.43</xmin>\n <ymin>0.46</ymin>\n <xmax>0.8</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Suit</name>\n <bndbox>\n <xmin>0.43</xmin>\n <ymin>0.74</ymin>\n <xmax>0.79</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/0561cd703efe784b.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Palm tree</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0</ymin>\n <xmax>0.6</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/042694c5469b641a.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.38</xmin>\n <ymin>0.63</ymin>\n <xmax>0.48</xmax>\n <ymax>0.95</ymax>\n </bndbox>\n </object>\n <object>\n <name>Drink</name>\n <bndbox>\n <xmin>0.04</xmin>\n <ymin>0.88</ymin>\n <xmax>0.07</xmax>\n <ymax>0.93</ymax>\n </bndbox>\n </object>\n <object>\n <name>Drink</name>\n <bndbox>\n <xmin>0.09</xmin>\n <ymin>0.94</ymin>\n <xmax>0.12</xmax>\n <ymax>0.96</ymax>\n </bndbox>\n </object>\n <object>\n <name>Drink</name>\n <bndbox>\n <xmin>0.09</xmin>\n <ymin>0.89</ymin>\n <xmax>0.12</xmax>\n <ymax>0.91</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/0394d5e00d63ae73.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0.47</ymin>\n <xmax>0.15</xmax>\n <ymax>0.59</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/04e06b6c528307fc.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.59</xmin>\n <ymin>0.15</ymin>\n <xmax>0.9</xmax>\n <ymax>0.99</ymax>\n </bndbox>\n </object>\n <object>\n <name>Suit</name>\n <bndbox>\n <xmin>0.59</xmin>\n <ymin>0.22</ymin>\n <xmax>0.9</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0.35</ymin>\n <xmax>0.1</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.8</xmin>\n <ymin>0.26</ymin>\n <xmax>0.87</xmax>\n <ymax>0.41</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human face</name>\n <bndbox>\n <xmin>0.67</xmin>\n <ymin>0.2</ymin>\n <xmax>0.75</xmax>\n <ymax>0.38</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/0168bdba83ff6aee.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Beer</name>\n <bndbox>\n <xmin>0.04</xmin>\n <ymin>0.49</ymin>\n <xmax>0.47</xmax>\n <ymax>0.89</ymax>\n </bndbox>\n </object>\n <object>\n <name>Beer</name>\n <bndbox>\n <xmin>0.51</xmin>\n <ymin>0.04</ymin>\n <xmax>0.92</xmax>\n <ymax>0.92</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/00c3e6527e715e60.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.05</xmin>\n <ymin>0.05</ymin>\n <xmax>0.22</xmax>\n <ymax>0.77</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.05</xmin>\n <ymin>0.08</ymin>\n <xmax>0.21</xmax>\n <ymax>0.7</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.28</xmin>\n <ymin>0.63</ymin>\n <xmax>0.33</xmax>\n <ymax>0.78</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.29</xmin>\n <ymin>0.51</ymin>\n <xmax>0.35</xmax>\n <ymax>0.63</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.37</xmin>\n <ymin>0.48</ymin>\n <xmax>0.42</xmax>\n <ymax>0.64</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.6</xmin>\n <ymin>0.52</ymin>\n <xmax>0.65</xmax>\n <ymax>0.66</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.66</xmin>\n <ymin>0.68</ymin>\n <xmax>0.76</xmax>\n <ymax>0.83</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.7</xmin>\n <ymin>0.53</ymin>\n <xmax>0.73</xmax>\n <ymax>0.63</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.79</xmin>\n <ymin>0.53</ymin>\n <xmax>0.86</xmax>\n <ymax>0.68</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/0706ab5b2a05504a.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Footwear</name>\n <bndbox>\n <xmin>0.8</xmin>\n <ymin>0.9</ymin>\n <xmax>0.91</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Footwear</name>\n <bndbox>\n <xmin>0.6</xmin>\n <ymin>0.92</ymin>\n <xmax>0.73</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Footwear</name>\n <bndbox>\n <xmin>0.62</xmin>\n <ymin>0.77</ymin>\n <xmax>0.76</xmax>\n <ymax>0.93</ymax>\n </bndbox>\n </object>\n <object>\n <name>Footwear</name>\n <bndbox>\n <xmin>0.62</xmin>\n <ymin>0.45</ymin>\n <xmax>0.72</xmax>\n <ymax>0.51</ymax>\n </bndbox>\n </object>\n <object>\n <name>Footwear</name>\n <bndbox>\n <xmin>0.87</xmin>\n <ymin>0.82</ymin>\n <xmax>1</xmax>\n <ymax>0.92</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/00035ecd5fa0418e.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Building</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0</ymin>\n <xmax>1</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/0788832e00f251b4.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Boy</name>\n <bndbox>\n <xmin>0.07</xmin>\n <ymin>0</ymin>\n <xmax>0.73</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Boy</name>\n <bndbox>\n <xmin>0.64</xmin>\n <ymin>0</ymin>\n <xmax>1</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Man</name>\n <bndbox>\n <xmin>0.06</xmin>\n <ymin>0.01</ymin>\n <xmax>0.74</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Man</name>\n <bndbox>\n <xmin>0.64</xmin>\n <ymin>0</ymin>\n <xmax>1</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0.57</ymin>\n <xmax>0.08</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.07</xmin>\n <ymin>0.44</ymin>\n <xmax>0.72</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.89</xmin>\n <ymin>0.6</ymin>\n <xmax>1</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human face</name>\n <bndbox>\n <xmin>0.64</xmin>\n <ymin>0</ymin>\n <xmax>0.89</xmax>\n <ymax>0.44</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/0894f7140ac3f570.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Laptop</name>\n <bndbox>\n <xmin>0.31</xmin>\n <ymin>0.07</ymin>\n <xmax>1</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0</ymin>\n <xmax>0.24</xmax>\n <ymax>0.33</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.91</xmin>\n <ymin>0.11</ymin>\n <xmax>1</xmax>\n <ymax>0.43</ymax>\n </bndbox>\n </object>\n <object>\n <name>Computer keyboard</name>\n <bndbox>\n <xmin>0.5</xmin>\n <ymin>0.58</ymin>\n <xmax>0.88</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/058fdfb8f40926ca.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Necklace</name>\n <bndbox>\n <xmin>0.15</xmin>\n <ymin>0.04</ymin>\n <xmax>0.78</xmax>\n <ymax>0.96</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/06fc1b177efd9f32.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Helicopter</name>\n <bndbox>\n <xmin>0.03</xmin>\n <ymin>0.16</ymin>\n <xmax>1</xmax>\n <ymax>0.83</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/01b321a558dfe242.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>House</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0.16</ymin>\n <xmax>0.98</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n <object>\n <name>Tree</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0</ymin>\n <xmax>0.67</xmax>\n <ymax>0.34</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0.6</ymin>\n <xmax>0.04</xmax>\n <ymax>0.7</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.04</xmin>\n <ymin>0.61</ymin>\n <xmax>0.1</xmax>\n <ymax>0.68</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.25</xmin>\n <ymin>0.61</ymin>\n <xmax>0.31</xmax>\n <ymax>0.73</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.33</xmin>\n <ymin>0.61</ymin>\n <xmax>0.4</xmax>\n <ymax>0.7</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.64</xmin>\n <ymin>0.6</ymin>\n <xmax>0.71</xmax>\n <ymax>0.68</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/05c138675a756be8.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.37</xmin>\n <ymin>0.36</ymin>\n <xmax>0.72</xmax>\n <ymax>0.97</ymax>\n </bndbox>\n </object>\n <object>\n <name>Woman</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0.13</ymin>\n <xmax>0.57</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Woman</name>\n <bndbox>\n <xmin>0.72</xmin>\n <ymin>0</ymin>\n <xmax>1</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0.44</ymin>\n <xmax>0.51</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.73</xmin>\n <ymin>0.32</ymin>\n <xmax>1</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human face</name>\n <bndbox>\n <xmin>0.22</xmin>\n <ymin>0.31</ymin>\n <xmax>0.26</xmax>\n <ymax>0.52</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human face</name>\n <bndbox>\n <xmin>0.49</xmin>\n <ymin>0.23</ymin>\n <xmax>0.58</xmax>\n <ymax>0.44</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/00e9d0f34286c4ec.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Table</name>\n <bndbox>\n <xmin>0.18</xmin>\n <ymin>0.52</ymin>\n <xmax>0.79</xmax>\n <ymax>0.92</ymax>\n </bndbox>\n </object>\n <object>\n <name>Man</name>\n <bndbox>\n <xmin>0.71</xmin>\n <ymin>0.44</ymin>\n <xmax>1</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/0035fc6e652c751a.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Airplane</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0.01</ymin>\n <xmax>1</xmax>\n <ymax>0.54</ymax>\n </bndbox>\n </object>\n <object>\n <name>Airplane</name>\n <bndbox>\n <xmin>0.21</xmin>\n <ymin>0.65</ymin>\n <xmax>0.77</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n <object>\n <name>Airplane</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.87</ymin>\n <xmax>0.61</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Airplane</name>\n <bndbox>\n <xmin>0.66</xmin>\n <ymin>0.82</ymin>\n <xmax>1</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Airplane</name>\n <bndbox>\n <xmin>0.79</xmin>\n <ymin>0.64</ymin>\n <xmax>1</xmax>\n <ymax>0.76</ymax>\n </bndbox>\n </object>\n <object>\n <name>Airplane</name>\n <bndbox>\n <xmin>0.79</xmin>\n <ymin>0.5</ymin>\n <xmax>1</xmax>\n <ymax>0.58</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/07c26c8545124bf1.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Rays and skates</name>\n <bndbox>\n <xmin>0.06</xmin>\n <ymin>0</ymin>\n <xmax>0.72</xmax>\n <ymax>0.81</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/01a4bfbb55917511.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Coin</name>\n <bndbox>\n <xmin>0.06</xmin>\n <ymin>0.09</ymin>\n <xmax>0.95</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/03657f20b2233b57.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Flower</name>\n <bndbox>\n <xmin>0.14</xmin>\n <ymin>0.17</ymin>\n <xmax>0.91</xmax>\n <ymax>0.93</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/082cc5a13dca4ab7.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.01</xmin>\n <ymin>0.04</ymin>\n <xmax>0.41</xmax>\n <ymax>0.49</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.47</xmin>\n <ymin>0.64</ymin>\n <xmax>1</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human face</name>\n <bndbox>\n <xmin>0.09</xmin>\n <ymin>0.3</ymin>\n <xmax>0.38</xmax>\n <ymax>0.75</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/025b3d1711760fab.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Woman</name>\n <bndbox>\n <xmin>0.01</xmin>\n <ymin>0</ymin>\n <xmax>0.97</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human face</name>\n <bndbox>\n <xmin>0.23</xmin>\n <ymin>0.23</ymin>\n <xmax>0.72</xmax>\n <ymax>0.59</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/02fce9f4529828c5.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Man</name>\n <bndbox>\n <xmin>0.81</xmin>\n <ymin>0.26</ymin>\n <xmax>0.97</xmax>\n <ymax>0.99</ymax>\n </bndbox>\n </object>\n <object>\n <name>Man</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0.3</ymin>\n <xmax>0.19</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Man</name>\n <bndbox>\n <xmin>0.2</xmin>\n <ymin>0.3</ymin>\n <xmax>0.35</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Man</name>\n <bndbox>\n <xmin>0.33</xmin>\n <ymin>0.34</ymin>\n <xmax>0.45</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/00a6ac1848c841fa.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Bread</name>\n <bndbox>\n <xmin>0.01</xmin>\n <ymin>0.07</ymin>\n <xmax>0.99</xmax>\n <ymax>0.99</ymax>\n </bndbox>\n </object>\n <object>\n <name>Fast food</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0.07</ymin>\n <xmax>1</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Bread</name>\n <bndbox>\n <xmin>0.52</xmin>\n <ymin>0.07</ymin>\n <xmax>1</xmax>\n <ymax>0.63</ymax>\n </bndbox>\n </object>\n <object>\n <name>Hamburger</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0.08</ymin>\n <xmax>1</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/05c77d7281394500.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Tree</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0</ymin>\n <xmax>1</xmax>\n <ymax>0.57</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/031d3808721990da.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Banana</name>\n <bndbox>\n <xmin>0.72</xmin>\n <ymin>0.49</ymin>\n <xmax>0.76</xmax>\n <ymax>0.52</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/0886b7c5598044d8.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Submarine sandwich</name>\n <bndbox>\n <xmin>0.24</xmin>\n <ymin>0.25</ymin>\n <xmax>0.88</xmax>\n <ymax>0.9</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/06818f73950c97a3.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Scorpion</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0</ymin>\n <xmax>0.91</xmax>\n <ymax>0.85</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/00026cbfecc8ec06.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Building</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0</ymin>\n <xmax>1</xmax>\n <ymax>0.82</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/000c885c11fefdba.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Human body</name>\n <bndbox>\n <xmin>0.02</xmin>\n <ymin>0.26</ymin>\n <xmax>0.97</xmax>\n <ymax>0.86</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human arm</name>\n <bndbox>\n <xmin>0.58</xmin>\n <ymin>0.62</ymin>\n <xmax>0.78</xmax>\n <ymax>0.86</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human arm</name>\n <bndbox>\n <xmin>0.58</xmin>\n <ymin>0.55</ymin>\n <xmax>0.7</xmax>\n <ymax>0.7</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/03b10c44ec450225.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Human face</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0.43</ymin>\n <xmax>0.07</xmax>\n <ymax>0.57</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human face</name>\n <bndbox>\n <xmin>0.1</xmin>\n <ymin>0.37</ymin>\n <xmax>0.13</xmax>\n <ymax>0.4</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human face</name>\n <bndbox>\n <xmin>0.1</xmin>\n <ymin>0.54</ymin>\n <xmax>0.12</xmax>\n <ymax>0.57</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human face</name>\n <bndbox>\n <xmin>0.39</xmin>\n <ymin>0.32</ymin>\n <xmax>0.41</xmax>\n <ymax>0.35</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human face</name>\n <bndbox>\n <xmin>0.98</xmin>\n <ymin>0.37</ymin>\n <xmax>1</xmax>\n <ymax>0.46</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/02e0f6706598bbc0.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Stairs</name>\n <bndbox>\n <xmin>0.41</xmin>\n <ymin>0.22</ymin>\n <xmax>0.63</xmax>\n <ymax>0.43</ymax>\n </bndbox>\n </object>\n <object>\n <name>Stairs</name>\n <bndbox>\n <xmin>0.74</xmin>\n <ymin>0.29</ymin>\n <xmax>1</xmax>\n <ymax>0.73</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/0673951a32e6d6cb.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Human face</name>\n <bndbox>\n <xmin>0.29</xmin>\n <ymin>0.38</ymin>\n <xmax>0.41</xmax>\n <ymax>0.6</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human face</name>\n <bndbox>\n <xmin>0.4</xmin>\n <ymin>0.16</ymin>\n <xmax>0.53</xmax>\n <ymax>0.4</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/033ba5f0c6d1488f.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Doll</name>\n <bndbox>\n <xmin>0.54</xmin>\n <ymin>0.05</ymin>\n <xmax>1</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/000f1c924e848a8d.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Window</name>\n <bndbox>\n <xmin>0.57</xmin>\n <ymin>0.76</ymin>\n <xmax>0.6</xmax>\n <ymax>0.79</ymax>\n </bndbox>\n </object>\n <object>\n <name>Window</name>\n <bndbox>\n <xmin>0.57</xmin>\n <ymin>0.72</ymin>\n <xmax>0.6</xmax>\n <ymax>0.75</ymax>\n </bndbox>\n </object>\n <object>\n <name>Window</name>\n <bndbox>\n <xmin>0.58</xmin>\n <ymin>0.22</ymin>\n <xmax>0.6</xmax>\n <ymax>0.24</ymax>\n </bndbox>\n </object>\n <object>\n <name>Window</name>\n <bndbox>\n <xmin>0.58</xmin>\n <ymin>0.67</ymin>\n <xmax>0.6</xmax>\n <ymax>0.7</ymax>\n </bndbox>\n </object>\n <object>\n <name>Window</name>\n <bndbox>\n <xmin>0.59</xmin>\n <ymin>0.17</ymin>\n <xmax>0.61</xmax>\n <ymax>0.19</ymax>\n </bndbox>\n </object>\n <object>\n <name>Window</name>\n <bndbox>\n <xmin>0.59</xmin>\n <ymin>0.95</ymin>\n <xmax>0.63</xmax>\n <ymax>0.98</ymax>\n </bndbox>\n </object>\n <object>\n <name>Window</name>\n <bndbox>\n <xmin>0.6</xmin>\n <ymin>0.87</ymin>\n <xmax>0.64</xmax>\n <ymax>0.89</ymax>\n </bndbox>\n </object>\n <object>\n <name>Window</name>\n <bndbox>\n <xmin>0.62</xmin>\n <ymin>0.26</ymin>\n <xmax>0.66</xmax>\n <ymax>0.31</ymax>\n </bndbox>\n </object>\n <object>\n <name>Window</name>\n <bndbox>\n <xmin>0.63</xmin>\n <ymin>0.24</ymin>\n <xmax>0.64</xmax>\n <ymax>0.24</ymax>\n </bndbox>\n </object>\n <object>\n <name>Window</name>\n <bndbox>\n <xmin>0.81</xmin>\n <ymin>0.67</ymin>\n <xmax>0.84</xmax>\n <ymax>0.7</ymax>\n </bndbox>\n </object>\n <object>\n <name>Window</name>\n <bndbox>\n <xmin>0.98</xmin>\n <ymin>0.97</ymin>\n <xmax>1</xmax>\n <ymax>0.99</ymax>\n </bndbox>\n </object>\n <object>\n <name>Window</name>\n <bndbox>\n <xmin>0.99</xmin>\n <ymin>0.92</ymin>\n <xmax>1</xmax>\n <ymax>0.96</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/0396726f3d15282c.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.45</xmin>\n <ymin>0.34</ymin>\n <xmax>0.58</xmax>\n <ymax>0.67</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.13</xmin>\n <ymin>0.38</ymin>\n <xmax>0.24</xmax>\n <ymax>0.89</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.13</xmin>\n <ymin>0.45</ymin>\n <xmax>0.22</xmax>\n <ymax>0.75</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/02e4ba3c19e0b891.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Bus</name>\n <bndbox>\n <xmin>0.03</xmin>\n <ymin>0.2</ymin>\n <xmax>0.99</xmax>\n <ymax>0.86</ymax>\n </bndbox>\n </object>\n <object>\n <name>Window</name>\n <bndbox>\n <xmin>0.05</xmin>\n <ymin>0.31</ymin>\n <xmax>0.55</xmax>\n <ymax>0.59</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/04a4d1b7a63f8bba.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Fountain</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0</ymin>\n <xmax>0.77</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/04f3d5a013c6b8aa.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Boot</name>\n <bndbox>\n <xmin>0.07</xmin>\n <ymin>0</ymin>\n <xmax>0.32</xmax>\n <ymax>0.74</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/01eb3b05cb59f523.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Tree</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0</ymin>\n <xmax>1</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/018d11ada92ef4af.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Door</name>\n <bndbox>\n <xmin>0.19</xmin>\n <ymin>0.24</ymin>\n <xmax>0.35</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Door</name>\n <bndbox>\n <xmin>0.74</xmin>\n <ymin>0.45</ymin>\n <xmax>0.81</xmax>\n <ymax>0.65</ymax>\n </bndbox>\n </object>\n <object>\n <name>Door</name>\n <bndbox>\n <xmin>0.81</xmin>\n <ymin>0.36</ymin>\n <xmax>0.96</xmax>\n <ymax>0.63</ymax>\n </bndbox>\n </object>\n <object>\n <name>Building</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0</ymin>\n <xmax>1</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/0093bb863d1b1311.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Carnivore</name>\n <bndbox>\n <xmin>0.47</xmin>\n <ymin>0.42</ymin>\n <xmax>0.64</xmax>\n <ymax>0.7</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/012818e7338cad4e.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Mushroom</name>\n <bndbox>\n <xmin>0.16</xmin>\n <ymin>0.11</ymin>\n <xmax>0.92</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/04864535f8d02b6d.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Fountain</name>\n <bndbox>\n <xmin>0.11</xmin>\n <ymin>0.51</ymin>\n <xmax>0.35</xmax>\n <ymax>0.63</ymax>\n </bndbox>\n </object>\n <object>\n <name>Fountain</name>\n <bndbox>\n <xmin>0.25</xmin>\n <ymin>0.32</ymin>\n <xmax>1</xmax>\n <ymax>0.64</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/0232e586ab9f76ff.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Human face</name>\n <bndbox>\n <xmin>0.38</xmin>\n <ymin>0.29</ymin>\n <xmax>0.48</xmax>\n <ymax>0.47</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/02cd474bae2c114e.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Human face</name>\n <bndbox>\n <xmin>0.45</xmin>\n <ymin>0.22</ymin>\n <xmax>0.75</xmax>\n <ymax>0.47</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/002ea5dc51eb3414.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Watercraft</name>\n <bndbox>\n <xmin>0.22</xmin>\n <ymin>0.34</ymin>\n <xmax>0.92</xmax>\n <ymax>0.51</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/05fd1531c4ec053b.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Bidet</name>\n <bndbox>\n <xmin>0.47</xmin>\n <ymin>0.62</ymin>\n <xmax>0.64</xmax>\n <ymax>0.78</ymax>\n </bndbox>\n </object>\n <object>\n <name>Toilet</name>\n <bndbox>\n <xmin>0.48</xmin>\n <ymin>0.61</ymin>\n <xmax>0.66</xmax>\n <ymax>0.78</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/00df21847a0b46b4.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Pig</name>\n <bndbox>\n <xmin>0.14</xmin>\n <ymin>0.31</ymin>\n <xmax>0.79</xmax>\n <ymax>0.68</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/08b8a2a25c1d0953.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Sports uniform</name>\n <bndbox>\n <xmin>0.22</xmin>\n <ymin>0.18</ymin>\n <xmax>0.77</xmax>\n <ymax>0.96</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/0178003482fa8d7f.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Tin can</name>\n <bndbox>\n <xmin>0.2</xmin>\n <ymin>0.13</ymin>\n <xmax>0.82</xmax>\n <ymax>0.9</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/002d4e7f08cf65bb.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Car</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0</ymin>\n <xmax>1</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/00a7bb9302f84145.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Bottle</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0.39</ymin>\n <xmax>0.4</xmax>\n <ymax>0.58</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/07607c1b34b524c2.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Human head</name>\n <bndbox>\n <xmin>0.43</xmin>\n <ymin>0.27</ymin>\n <xmax>0.56</xmax>\n <ymax>0.48</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human body</name>\n <bndbox>\n <xmin>0.14</xmin>\n <ymin>0.1</ymin>\n <xmax>0.28</xmax>\n <ymax>0.68</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human body</name>\n <bndbox>\n <xmin>0.3</xmin>\n <ymin>0.09</ymin>\n <xmax>0.45</xmax>\n <ymax>0.61</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human leg</name>\n <bndbox>\n <xmin>0.18</xmin>\n <ymin>0.38</ymin>\n <xmax>0.21</xmax>\n <ymax>0.68</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human leg</name>\n <bndbox>\n <xmin>0.21</xmin>\n <ymin>0.39</ymin>\n <xmax>0.32</xmax>\n <ymax>0.68</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human head</name>\n <bndbox>\n <xmin>0.44</xmin>\n <ymin>0.07</ymin>\n <xmax>0.52</xmax>\n <ymax>0.14</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human head</name>\n <bndbox>\n <xmin>0.79</xmin>\n <ymin>0.14</ymin>\n <xmax>0.82</xmax>\n <ymax>0.18</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human head</name>\n <bndbox>\n <xmin>0.82</xmin>\n <ymin>0.12</ymin>\n <xmax>0.85</xmax>\n <ymax>0.16</ymax>\n </bndbox>\n </object>\n <object>\n <name>Man</name>\n <bndbox>\n <xmin>0.59</xmin>\n <ymin>0.11</ymin>\n <xmax>0.77</xmax>\n <ymax>0.66</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.14</xmin>\n <ymin>0.15</ymin>\n <xmax>0.3</xmax>\n <ymax>0.67</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human hand</name>\n <bndbox>\n <xmin>0.29</xmin>\n <ymin>0.23</ymin>\n <xmax>0.33</xmax>\n <ymax>0.28</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/01bedaf614b4f4fc.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Skyscraper</name>\n <bndbox>\n <xmin>0.29</xmin>\n <ymin>0.19</ymin>\n <xmax>0.45</xmax>\n <ymax>0.65</ymax>\n </bndbox>\n </object>\n <object>\n <name>Taxi</name>\n <bndbox>\n <xmin>0.77</xmin>\n <ymin>0.84</ymin>\n <xmax>0.92</xmax>\n <ymax>0.99</ymax>\n </bndbox>\n </object>\n <object>\n <name>Car</name>\n <bndbox>\n <xmin>0.18</xmin>\n <ymin>0.66</ymin>\n <xmax>0.19</xmax>\n <ymax>0.67</ymax>\n </bndbox>\n </object>\n <object>\n <name>Car</name>\n <bndbox>\n <xmin>0.19</xmin>\n <ymin>0.66</ymin>\n <xmax>0.2</xmax>\n <ymax>0.67</ymax>\n </bndbox>\n </object>\n <object>\n <name>Car</name>\n <bndbox>\n <xmin>0.2</xmin>\n <ymin>0.67</ymin>\n <xmax>0.21</xmax>\n <ymax>0.68</ymax>\n </bndbox>\n </object>\n <object>\n <name>Car</name>\n <bndbox>\n <xmin>0.75</xmin>\n <ymin>0.83</ymin>\n <xmax>0.92</xmax>\n <ymax>0.97</ymax>\n </bndbox>\n </object>\n <object>\n <name>Car</name>\n <bndbox>\n <xmin>0.87</xmin>\n <ymin>0.89</ymin>\n <xmax>0.98</xmax>\n <ymax>0.98</ymax>\n </bndbox>\n </object>\n <object>\n <name>Taxi</name>\n <bndbox>\n <xmin>0.19</xmin>\n <ymin>0.66</ymin>\n <xmax>0.21</xmax>\n <ymax>0.67</ymax>\n </bndbox>\n </object>\n <object>\n <name>Taxi</name>\n <bndbox>\n <xmin>0.21</xmin>\n <ymin>0.68</ymin>\n <xmax>0.23</xmax>\n <ymax>0.7</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/019aba9c37a0d147.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.41</xmin>\n <ymin>0.43</ymin>\n <xmax>0.72</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.68</xmin>\n <ymin>0.49</ymin>\n <xmax>0.75</xmax>\n <ymax>0.7</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/07f3248b1938faab.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.11</xmin>\n <ymin>0.36</ymin>\n <xmax>0.56</xmax>\n <ymax>0.98</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/06ba7164fcdc8af4.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Tire</name>\n <bndbox>\n <xmin>0.23</xmin>\n <ymin>0.06</ymin>\n <xmax>0.8</xmax>\n <ymax>0.79</ymax>\n </bndbox>\n </object>\n <object>\n <name>Tire</name>\n <bndbox>\n <xmin>0.84</xmin>\n <ymin>0.49</ymin>\n <xmax>1</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/03b6e6e005af384e.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Poster</name>\n <bndbox>\n <xmin>0.03</xmin>\n <ymin>0.01</ymin>\n <xmax>1</xmax>\n <ymax>0.99</ymax>\n </bndbox>\n </object>\n <object>\n <name>Skateboard</name>\n <bndbox>\n <xmin>0.81</xmin>\n <ymin>0.53</ymin>\n <xmax>1</xmax>\n <ymax>0.69</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.75</xmin>\n <ymin>0.14</ymin>\n <xmax>1</xmax>\n <ymax>0.54</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/078f7b72c63c50f2.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Convenience store</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0.01</ymin>\n <xmax>1</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/00a1a0eddc2f0200.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.47</xmin>\n <ymin>0.55</ymin>\n <xmax>0.58</xmax>\n <ymax>0.93</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.02</xmin>\n <ymin>0.37</ymin>\n <xmax>0.12</xmax>\n <ymax>0.74</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.22</xmin>\n <ymin>0.13</ymin>\n <xmax>0.31</xmax>\n <ymax>0.36</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.26</xmin>\n <ymin>0.53</ymin>\n <xmax>0.38</xmax>\n <ymax>0.93</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.34</xmin>\n <ymin>0.1</ymin>\n <xmax>0.4</xmax>\n <ymax>0.27</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.34</xmin>\n <ymin>0.19</ymin>\n <xmax>0.42</xmax>\n <ymax>0.59</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.41</xmin>\n <ymin>0.11</ymin>\n <xmax>0.51</xmax>\n <ymax>0.35</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.48</xmin>\n <ymin>0.53</ymin>\n <xmax>0.57</xmax>\n <ymax>0.95</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.51</xmin>\n <ymin>0.1</ymin>\n <xmax>0.57</xmax>\n <ymax>0.33</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.62</xmin>\n <ymin>0.08</ymin>\n <xmax>0.68</xmax>\n <ymax>0.34</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.72</xmin>\n <ymin>0.08</ymin>\n <xmax>0.79</xmax>\n <ymax>0.33</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.72</xmin>\n <ymin>0.59</ymin>\n <xmax>0.82</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.82</xmin>\n <ymin>0.1</ymin>\n <xmax>0.87</xmax>\n <ymax>0.28</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.92</xmin>\n <ymin>0.11</ymin>\n <xmax>1</xmax>\n <ymax>0.29</ymax>\n </bndbox>\n </object>\n <object>\n <name>Musical keyboard</name>\n <bndbox>\n <xmin>0.23</xmin>\n <ymin>0.34</ymin>\n <xmax>0.44</xmax>\n <ymax>0.46</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.03</xmin>\n <ymin>0.4</ymin>\n <xmax>0.12</xmax>\n <ymax>0.74</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.23</xmin>\n <ymin>0.16</ymin>\n <xmax>0.32</xmax>\n <ymax>0.35</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.27</xmin>\n <ymin>0.58</ymin>\n <xmax>0.38</xmax>\n <ymax>0.91</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.33</xmin>\n <ymin>0.14</ymin>\n <xmax>0.41</xmax>\n <ymax>0.3</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.36</xmin>\n <ymin>0.25</ymin>\n <xmax>0.42</xmax>\n <ymax>0.56</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.43</xmin>\n <ymin>0.16</ymin>\n <xmax>0.49</xmax>\n <ymax>0.31</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.5</xmin>\n <ymin>0.14</ymin>\n <xmax>0.57</xmax>\n <ymax>0.3</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.62</xmin>\n <ymin>0.15</ymin>\n <xmax>0.69</xmax>\n <ymax>0.35</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.74</xmin>\n <ymin>0.64</ymin>\n <xmax>0.84</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.74</xmin>\n <ymin>0.24</ymin>\n <xmax>0.8</xmax>\n <ymax>0.42</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.82</xmin>\n <ymin>0.15</ymin>\n <xmax>0.9</xmax>\n <ymax>0.29</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.94</xmin>\n <ymin>0.16</ymin>\n <xmax>0.99</xmax>\n <ymax>0.26</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.94</xmin>\n <ymin>0.27</ymin>\n <xmax>1</xmax>\n <ymax>0.45</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/02c2ee0d1cdfa3f8.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0.53</ymin>\n <xmax>0.12</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.78</xmin>\n <ymin>0.38</ymin>\n <xmax>0.99</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/073a724b36613378.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Butterfly</name>\n <bndbox>\n <xmin>0.06</xmin>\n <ymin>0.01</ymin>\n <xmax>0.82</xmax>\n <ymax>0.98</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/0084fe212a0eacc1.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Footwear</name>\n <bndbox>\n <xmin>0.69</xmin>\n <ymin>0.91</ymin>\n <xmax>0.85</xmax>\n <ymax>0.97</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human face</name>\n <bndbox>\n <xmin>0.29</xmin>\n <ymin>0.42</ymin>\n <xmax>0.36</xmax>\n <ymax>0.49</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human face</name>\n <bndbox>\n <xmin>0.57</xmin>\n <ymin>0.49</ymin>\n <xmax>0.62</xmax>\n <ymax>0.53</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human face</name>\n <bndbox>\n <xmin>0.71</xmin>\n <ymin>0.37</ymin>\n <xmax>0.8</xmax>\n <ymax>0.45</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/01e809f6c07375fa.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Vehicle</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.57</ymin>\n <xmax>0.16</xmax>\n <ymax>0.66</ymax>\n </bndbox>\n </object>\n <object>\n <name>Vehicle</name>\n <bndbox>\n <xmin>0.17</xmin>\n <ymin>0.59</ymin>\n <xmax>0.23</xmax>\n <ymax>0.66</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/006c7d7302b1863c.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Jeans</name>\n <bndbox>\n <xmin>0.46</xmin>\n <ymin>0.81</ymin>\n <xmax>0.89</xmax>\n <ymax>0.85</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/058743e57e54ff6d.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Chair</name>\n <bndbox>\n <xmin>0.56</xmin>\n <ymin>0.65</ymin>\n <xmax>0.59</xmax>\n <ymax>0.79</ymax>\n </bndbox>\n </object>\n <object>\n <name>Man</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0.49</ymin>\n <xmax>0.18</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Man</name>\n <bndbox>\n <xmin>0.24</xmin>\n <ymin>0.5</ymin>\n <xmax>0.69</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Man</name>\n <bndbox>\n <xmin>0.94</xmin>\n <ymin>0.23</ymin>\n <xmax>1</xmax>\n <ymax>0.43</ymax>\n </bndbox>\n </object>\n <object>\n <name>Footwear</name>\n <bndbox>\n <xmin>0.77</xmin>\n <ymin>0.83</ymin>\n <xmax>0.83</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/083ae5cfc85f79c7.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Cookie</name>\n <bndbox>\n <xmin>0.06</xmin>\n <ymin>0.37</ymin>\n <xmax>0.96</xmax>\n <ymax>0.82</ymax>\n </bndbox>\n </object>\n <object>\n <name>Lemon</name>\n <bndbox>\n <xmin>0.19</xmin>\n <ymin>0.01</ymin>\n <xmax>0.5</xmax>\n <ymax>0.16</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/00c7f84d61adce05.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Bottle</name>\n <bndbox>\n <xmin>0.45</xmin>\n <ymin>0.05</ymin>\n <xmax>0.94</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/04fe3fa6663e8b13.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Pumpkin</name>\n <bndbox>\n <xmin>0.29</xmin>\n <ymin>0.45</ymin>\n <xmax>0.5</xmax>\n <ymax>0.73</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/0838be59f01f3a25.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Traffic light</name>\n <bndbox>\n <xmin>0.58</xmin>\n <ymin>0.12</ymin>\n <xmax>0.63</xmax>\n <ymax>0.31</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/05b775285c85478a.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Rifle</name>\n <bndbox>\n <xmin>0.61</xmin>\n <ymin>0.48</ymin>\n <xmax>0.99</xmax>\n <ymax>0.73</ymax>\n </bndbox>\n </object>\n <object>\n <name>Shotgun</name>\n <bndbox>\n <xmin>0.61</xmin>\n <ymin>0.47</ymin>\n <xmax>1</xmax>\n <ymax>0.71</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/01bf90ca6b7fa09c.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Moths and butterflies</name>\n <bndbox>\n <xmin>0.27</xmin>\n <ymin>0.31</ymin>\n <xmax>0.82</xmax>\n <ymax>0.6</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/06d3a3d570ff9566.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Footwear</name>\n <bndbox>\n <xmin>0.23</xmin>\n <ymin>0.86</ymin>\n <xmax>0.31</xmax>\n <ymax>0.98</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/05bb4bd3c8144918.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Picture frame</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0.09</ymin>\n <xmax>0.08</xmax>\n <ymax>0.25</ymax>\n </bndbox>\n </object>\n <object>\n <name>Picture frame</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0.28</ymin>\n <xmax>0.11</xmax>\n <ymax>0.46</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human face</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.12</ymin>\n <xmax>0.88</xmax>\n <ymax>0.38</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/06e4716d42f540f4.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Dress</name>\n <bndbox>\n <xmin>0.22</xmin>\n <ymin>0.23</ymin>\n <xmax>0.84</xmax>\n <ymax>0.97</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/065f1282978e2ea7.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.2</xmin>\n <ymin>0.23</ymin>\n <xmax>0.42</xmax>\n <ymax>0.89</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.59</xmin>\n <ymin>0.29</ymin>\n <xmax>0.88</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/02241ed7687a730f.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Lizard</name>\n <bndbox>\n <xmin>0.01</xmin>\n <ymin>0.38</ymin>\n <xmax>0.85</xmax>\n <ymax>0.61</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/08a11374e4773407.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Wok</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0</ymin>\n <xmax>0.36</xmax>\n <ymax>0.29</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/04bcef396286edaf.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>House</name>\n <bndbox>\n <xmin>0.06</xmin>\n <ymin>0</ymin>\n <xmax>0.93</xmax>\n <ymax>0.93</ymax>\n </bndbox>\n </object>\n <object>\n <name>Car</name>\n <bndbox>\n <xmin>0.39</xmin>\n <ymin>0.7</ymin>\n <xmax>0.93</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/07877978a7751105.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Car</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0</ymin>\n <xmax>0.79</xmax>\n <ymax>0.39</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/082e065e93eca3aa.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.19</xmin>\n <ymin>0.52</ymin>\n <xmax>0.26</xmax>\n <ymax>0.81</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0.97</xmin>\n <ymin>0.66</ymin>\n <xmax>1</xmax>\n <ymax>0.82</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/048dc433a4dc4bf4.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Land vehicle</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0.01</ymin>\n <xmax>0.97</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/00d9cb4ad4a99a22.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.66</xmin>\n <ymin>0.3</ymin>\n <xmax>0.83</xmax>\n <ymax>0.86</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.01</xmin>\n <ymin>0.36</ymin>\n <xmax>0.07</xmax>\n <ymax>0.91</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.02</xmin>\n <ymin>0.28</ymin>\n <xmax>0.22</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.34</xmin>\n <ymin>0.27</ymin>\n <xmax>0.5</xmax>\n <ymax>0.8</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.77</xmin>\n <ymin>0.12</ymin>\n <xmax>1</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/06d1f01bc2ddf996.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Tree</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0</ymin>\n <xmax>0.47</xmax>\n <ymax>0.64</ymax>\n </bndbox>\n </object>\n <object>\n <name>Tree</name>\n <bndbox>\n <xmin>0.51</xmin>\n <ymin>0.19</ymin>\n <xmax>0.63</xmax>\n <ymax>0.25</ymax>\n </bndbox>\n </object>\n <object>\n <name>Tree</name>\n <bndbox>\n <xmin>0.73</xmin>\n <ymin>0</ymin>\n <xmax>1</xmax>\n <ymax>0.53</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/08588a4386d8e18b.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Musical keyboard</name>\n <bndbox>\n <xmin>0.52</xmin>\n <ymin>0.47</ymin>\n <xmax>0.72</xmax>\n <ymax>0.83</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/022f6d114a323c19.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Human face</name>\n <bndbox>\n <xmin>0.11</xmin>\n <ymin>0.31</ymin>\n <xmax>0.16</xmax>\n <ymax>0.42</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human face</name>\n <bndbox>\n <xmin>0.24</xmin>\n <ymin>0.3</ymin>\n <xmax>0.29</xmax>\n <ymax>0.4</ymax>\n </bndbox>\n </object>\n <object>\n <name>Human face</name>\n <bndbox>\n <xmin>0.98</xmin>\n <ymin>0.22</ymin>\n <xmax>1</xmax>\n <ymax>0.31</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/04366bee2e561c25.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Human face</name>\n <bndbox>\n <xmin>0.39</xmin>\n <ymin>0.17</ymin>\n <xmax>0.51</xmax>\n <ymax>0.47</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/004570a9c185f956.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Human body</name>\n <bndbox>\n <xmin>0.16</xmin>\n <ymin>0.07</ymin>\n <xmax>0.84</xmax>\n <ymax>0.75</ymax>\n </bndbox>\n </object>\n <object>\n <name>Microphone</name>\n <bndbox>\n <xmin>0.53</xmin>\n <ymin>0.16</ymin>\n <xmax>0.76</xmax>\n <ymax>0.21</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/03ffdae09ea744b1.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Dessert</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0.13</ymin>\n <xmax>0.49</xmax>\n <ymax>0.93</ymax>\n </bndbox>\n </object>\n <object>\n <name>Pastry</name>\n <bndbox>\n <xmin>0.34</xmin>\n <ymin>0.1</ymin>\n <xmax>0.56</xmax>\n <ymax>0.39</ymax>\n </bndbox>\n </object>\n <object>\n <name>Pastry</name>\n <bndbox>\n <xmin>0.46</xmin>\n <ymin>0.26</ymin>\n <xmax>0.83</xmax>\n <ymax>0.81</ymax>\n </bndbox>\n </object>\n <object>\n <name>Pastry</name>\n <bndbox>\n <xmin>0.47</xmin>\n <ymin>0.06</ymin>\n <xmax>0.69</xmax>\n <ymax>0.27</ymax>\n </bndbox>\n </object>\n <object>\n <name>Pastry</name>\n <bndbox>\n <xmin>0.61</xmin>\n <ymin>0.21</ymin>\n <xmax>0.96</xmax>\n <ymax>0.69</ymax>\n </bndbox>\n </object>\n <object>\n <name>Pastry</name>\n <bndbox>\n <xmin>0.62</xmin>\n <ymin>0.03</ymin>\n <xmax>1</xmax>\n <ymax>0.56</ymax>\n </bndbox>\n </object>\n <object>\n <name>Pastry</name>\n <bndbox>\n <xmin>0.72</xmin>\n <ymin>0.18</ymin>\n <xmax>1</xmax>\n <ymax>0.59</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/020c79b7040bd12a.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Butterfly</name>\n <bndbox>\n <xmin>0.29</xmin>\n <ymin>0.19</ymin>\n <xmax>1</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/007d9ec191c44dfb.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Cat</name>\n <bndbox>\n <xmin>0.06</xmin>\n <ymin>0</ymin>\n <xmax>0.84</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/060f03ccc21bcaba.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Human face</name>\n <bndbox>\n <xmin>0.22</xmin>\n <ymin>0.1</ymin>\n <xmax>0.78</xmax>\n <ymax>0.63</ymax>\n </bndbox>\n </object>\n <object>\n <name>Man</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0.04</ymin>\n <xmax>1</xmax>\n <ymax>1</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/01ba91fb1c069f5d.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Stretcher</name>\n <bndbox>\n <xmin>0.47</xmin>\n <ymin>0.46</ymin>\n <xmax>0.88</xmax>\n <ymax>0.78</ymax>\n </bndbox>\n </object>\n <object>\n <name>Person</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0.21</ymin>\n <xmax>0.12</xmax>\n <ymax>0.75</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0.26</ymin>\n <xmax>0.08</xmax>\n <ymax>0.7</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.16</xmin>\n <ymin>0.23</ymin>\n <xmax>0.27</xmax>\n <ymax>0.65</ymax>\n </bndbox>\n </object>\n <object>\n <name>Clothing</name>\n <bndbox>\n <xmin>0.26</xmin>\n <ymin>0.26</ymin>\n <xmax>0.47</xmax>\n <ymax>0.93</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/011186a4555d9d1b.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Animal</name>\n <bndbox>\n <xmin>0.33</xmin>\n <ymin>0.42</ymin>\n <xmax>0.86</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
[
{
"content": [
{
"path": null,
"text": "You are an AI vision assistant specialized in object detection.\n\nThe user will provide exactly one image. \nYour task is to detect all visible, unoccluded objects in that image and output their categories and bounding boxes.\n\nOnly include objects that are **fully visible** — \ndo not annotate any object that is partially hidden, truncated, or occluded by other objects.\n\nFor each detected object, include:\n- <name>: the object category (e.g., \"person\", \"car\", \"dog\").\n- <bndbox>: the bounding box coordinates, using:\n <xmin>, <ymin>, <xmax>, <ymax>\n\nCoordinates must be **normalized** to the range [0, 1], \nrelative to the image width and height:\n- xmin = left / image_width \n- ymin = top / image_height \n- xmax = right / image_width \n- ymax = bottom / image_height\n\nYou must output **only** the results in the following **XML format**, with no explanations, no comments, and no extra text:\n\n```xml\n<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>person</name>\n <bndbox>\n <xmin>0.12</xmin>\n <ymin>0.08</ymin>\n <xmax>0.45</xmax>\n <ymax>0.72</ymax>\n </bndbox>\n </object>\n <object>\n <name>dog</name>\n <bndbox>\n <xmin>0.55</xmin>\n <ymin>0.35</ymin>\n <xmax>0.80</xmax>\n <ymax>0.88</ymax>\n </bndbox>\n </object>\n</annotation>\n```\n",
"type": "text"
}
],
"role": "system"
},
{
"content": [
{
"path": "/mnt/shared/p02/alex/fiftyone/open-images-v7/train/data/05853e14f6ec66c5.jpg",
"text": null,
"type": "image"
}
],
"role": "user"
},
{
"content": [
{
"path": null,
"text": "<?xml version=\"1.0\" ?>\n<annotation>\n <object>\n <name>Plate</name>\n <bndbox>\n <xmin>0</xmin>\n <ymin>0</ymin>\n <xmax>0.23</xmax>\n <ymax>0.4</ymax>\n </bndbox>\n </object>\n <object>\n <name>Plate</name>\n <bndbox>\n <xmin>0.21</xmin>\n <ymin>0.09</ymin>\n <xmax>0.77</xmax>\n <ymax>0.89</ymax>\n </bndbox>\n </object>\n <object>\n <name>Plate</name>\n <bndbox>\n <xmin>0.86</xmin>\n <ymin>0</ymin>\n <xmax>1</xmax>\n <ymax>0.65</ymax>\n </bndbox>\n </object>\n</annotation>",
"type": "text"
}
],
"role": "assistant"
}
] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.