Datasets:
Add text description
Browse files
CC6204-Hackaton-Cub-Dataset.py
CHANGED
@@ -98,19 +98,24 @@ class CubDataset(datasets.GeneratorBasedBuilder):
|
|
98 |
def _split_generators(self, dl_manager):
|
99 |
train_files = []
|
100 |
train_idx = []
|
|
|
101 |
test_files = []
|
102 |
test_idx = []
|
103 |
|
104 |
# Download images
|
105 |
-
|
106 |
-
|
107 |
-
|
|
|
|
|
|
|
|
|
108 |
img_idx = _IMGNAME2ID[os.path.basename(img)]
|
109 |
if img_idx in _TRAIN_IDX_SET:
|
110 |
-
train_files.append(img)
|
111 |
train_idx.append(img_idx)
|
112 |
else:
|
113 |
-
test_files.append(img)
|
114 |
test_idx.append(img_idx)
|
115 |
|
116 |
return [
|
@@ -134,11 +139,11 @@ class CubDataset(datasets.GeneratorBasedBuilder):
|
|
134 |
def _generate_examples(self, files, image_idx):
|
135 |
|
136 |
for i, path in enumerate(files):
|
137 |
-
file_name = os.path.basename(path)
|
138 |
if file_name.endswith(".jpg"):
|
139 |
yield i, {
|
140 |
-
"image": path,
|
141 |
-
"description":
|
142 |
"label": _ID2LABEL[_IMGID2CLASSID[image_idx[i]]],
|
143 |
}
|
144 |
|
|
|
98 |
def _split_generators(self, dl_manager):
|
99 |
train_files = []
|
100 |
train_idx = []
|
101 |
+
|
102 |
test_files = []
|
103 |
test_idx = []
|
104 |
|
105 |
# Download images
|
106 |
+
img_data_files = dl_manager.download_and_extract(_URLS["image_urls"])
|
107 |
+
text_data_files = dl_manager.download_and_extract(_URLS["text_urls"])
|
108 |
+
|
109 |
+
img_path_files = dl_manager.iter_files(img_data_files)
|
110 |
+
text_path_files = dl_manager.iter_files(text_data_files)
|
111 |
+
|
112 |
+
for img, text in zip(img_path_files, text_path_files):
|
113 |
img_idx = _IMGNAME2ID[os.path.basename(img)]
|
114 |
if img_idx in _TRAIN_IDX_SET:
|
115 |
+
train_files.append((img, text))
|
116 |
train_idx.append(img_idx)
|
117 |
else:
|
118 |
+
test_files.append((img, text))
|
119 |
test_idx.append(img_idx)
|
120 |
|
121 |
return [
|
|
|
139 |
def _generate_examples(self, files, image_idx):
|
140 |
|
141 |
for i, path in enumerate(files):
|
142 |
+
file_name = os.path.basename(path[0])
|
143 |
if file_name.endswith(".jpg"):
|
144 |
yield i, {
|
145 |
+
"image": path[0],
|
146 |
+
"description": path[1],
|
147 |
"label": _ID2LABEL[_IMGID2CLASSID[image_idx[i]]],
|
148 |
}
|
149 |
|