Spaces:
Runtime error
Runtime error
Enhance accuracy
Browse files- app.py +50 -35
- examples/bears.jpg +0 -0
- examples/cats.jpg +0 -0
- examples/fish.jpg +0 -0
app.py
CHANGED
@@ -12,13 +12,13 @@ import PIL
|
|
12 |
import torch
|
13 |
from segment_anything import SamAutomaticMaskGenerator, sam_model_registry
|
14 |
|
15 |
-
|
16 |
CHECKPOINT_PATH = os.path.join(os.path.expanduser("~"), ".cache", "SAM")
|
17 |
CHECKPOINT_NAME = "sam_vit_h_4b8939.pth"
|
18 |
CHECKPOINT_URL = "https://dl.fbaipublicfiles.com/segment_anything/sam_vit_h_4b8939.pth"
|
19 |
MODEL_TYPE = "default"
|
20 |
-
MAX_WIDTH = MAX_HEIGHT =
|
21 |
-
|
|
|
22 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
23 |
|
24 |
|
@@ -55,23 +55,19 @@ def adjust_image_size(image: np.ndarray) -> np.ndarray:
|
|
55 |
|
56 |
|
57 |
@torch.no_grad()
|
58 |
-
def
|
59 |
model, preprocess = load_clip()
|
60 |
-
preprocessed =
|
61 |
-
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
-
img_features /= img_features.norm(dim=-1, keepdim=True)
|
66 |
-
txt_features /= txt_features.norm(dim=-1, keepdim=True)
|
67 |
-
similarity = (100 * img_features @ txt_features.T).softmax(0)
|
68 |
-
return similarity
|
69 |
|
70 |
|
71 |
def crop_image(image: np.ndarray, mask: Dict[str, Any]) -> PIL.Image.Image:
|
72 |
x, y, w, h = mask["bbox"]
|
73 |
masked = image * np.expand_dims(mask["segmentation"], -1)
|
74 |
-
crop = masked[y: y + h, x: x + w]
|
75 |
if h > w:
|
76 |
top, bottom, left, right = 0, 0, (h - w) // 2, (h - w) // 2
|
77 |
else:
|
@@ -86,11 +82,14 @@ def crop_image(image: np.ndarray, mask: Dict[str, Any]) -> PIL.Image.Image:
|
|
86 |
cv2.BORDER_CONSTANT,
|
87 |
value=(0, 0, 0),
|
88 |
)
|
89 |
-
crop = cv2.cvtColor(crop, cv2.COLOR_BGR2RGB)
|
90 |
crop = PIL.Image.fromarray(crop)
|
91 |
return crop
|
92 |
|
93 |
|
|
|
|
|
|
|
|
|
94 |
def filter_masks(
|
95 |
image: np.ndarray,
|
96 |
masks: List[Dict[str, Any]],
|
@@ -99,26 +98,19 @@ def filter_masks(
|
|
99 |
query: str,
|
100 |
clip_threshold: float,
|
101 |
) -> List[Dict[str, Any]]:
|
102 |
-
cropped_masks: List[PIL.Image.Image] = []
|
103 |
filtered_masks: List[Dict[str, Any]] = []
|
104 |
|
105 |
-
for mask in masks:
|
106 |
if (
|
107 |
mask["predicted_iou"] < predicted_iou_threshold
|
108 |
or mask["stability_score"] < stability_score_threshold
|
109 |
or image.shape[:2] != mask["segmentation"].shape[:2]
|
|
|
|
|
110 |
):
|
111 |
continue
|
112 |
-
filtered_masks.append(mask)
|
113 |
-
cropped_masks.append(crop_image(image, mask))
|
114 |
|
115 |
-
|
116 |
-
scores = get_scores(cropped_masks, query)
|
117 |
-
filtered_masks = [
|
118 |
-
filtered_masks[i]
|
119 |
-
for i, score in enumerate(scores)
|
120 |
-
if score > clip_threshold
|
121 |
-
]
|
122 |
|
123 |
return filtered_masks
|
124 |
|
@@ -140,7 +132,7 @@ def draw_masks(
|
|
140 |
contours, _ = cv2.findContours(
|
141 |
np.uint8(mask["segmentation"]), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE
|
142 |
)
|
143 |
-
cv2.drawContours(image, contours, -1, (
|
144 |
return image
|
145 |
|
146 |
|
@@ -152,8 +144,11 @@ def segment(
|
|
152 |
query: str,
|
153 |
) -> PIL.ImageFile.ImageFile:
|
154 |
mask_generator = load_mask_generator()
|
|
|
|
|
|
|
155 |
# reduce the size to save gpu memory
|
156 |
-
image = adjust_image_size(
|
157 |
masks = mask_generator.generate(image)
|
158 |
masks = filter_masks(
|
159 |
image,
|
@@ -164,7 +159,6 @@ def segment(
|
|
164 |
clip_threshold,
|
165 |
)
|
166 |
image = draw_masks(image, masks)
|
167 |
-
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
|
168 |
image = PIL.Image.fromarray(image)
|
169 |
return image
|
170 |
|
@@ -185,31 +179,52 @@ demo = gr.Interface(
|
|
185 |
[
|
186 |
0.9,
|
187 |
0.8,
|
188 |
-
0.
|
189 |
os.path.join(os.path.dirname(__file__), "examples/dog.jpg"),
|
190 |
-
"
|
191 |
],
|
192 |
[
|
193 |
0.9,
|
194 |
0.8,
|
195 |
-
0.
|
196 |
os.path.join(os.path.dirname(__file__), "examples/city.jpg"),
|
197 |
"building",
|
198 |
],
|
199 |
[
|
200 |
0.9,
|
201 |
0.8,
|
202 |
-
0.
|
203 |
os.path.join(os.path.dirname(__file__), "examples/food.jpg"),
|
204 |
-
"
|
205 |
],
|
206 |
[
|
207 |
0.9,
|
208 |
0.8,
|
209 |
-
0.
|
210 |
os.path.join(os.path.dirname(__file__), "examples/horse.jpg"),
|
211 |
"horse",
|
212 |
],
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
213 |
],
|
214 |
)
|
215 |
|
|
|
12 |
import torch
|
13 |
from segment_anything import SamAutomaticMaskGenerator, sam_model_registry
|
14 |
|
|
|
15 |
CHECKPOINT_PATH = os.path.join(os.path.expanduser("~"), ".cache", "SAM")
|
16 |
CHECKPOINT_NAME = "sam_vit_h_4b8939.pth"
|
17 |
CHECKPOINT_URL = "https://dl.fbaipublicfiles.com/segment_anything/sam_vit_h_4b8939.pth"
|
18 |
MODEL_TYPE = "default"
|
19 |
+
MAX_WIDTH = MAX_HEIGHT = 1024
|
20 |
+
TOP_K_OBJ = 100
|
21 |
+
THRESHOLD = 0.85
|
22 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
23 |
|
24 |
|
|
|
55 |
|
56 |
|
57 |
@torch.no_grad()
|
58 |
+
def get_score(crop: PIL.Image.Image, texts: List[str]) -> torch.Tensor:
|
59 |
model, preprocess = load_clip()
|
60 |
+
preprocessed = preprocess(crop).unsqueeze(0).to(device)
|
61 |
+
tokens = clip.tokenize(texts).to(device)
|
62 |
+
logits_per_image, _ = model(preprocessed, tokens)
|
63 |
+
similarity = logits_per_image.softmax(-1).cpu()
|
64 |
+
return similarity[0, 0]
|
|
|
|
|
|
|
|
|
65 |
|
66 |
|
67 |
def crop_image(image: np.ndarray, mask: Dict[str, Any]) -> PIL.Image.Image:
|
68 |
x, y, w, h = mask["bbox"]
|
69 |
masked = image * np.expand_dims(mask["segmentation"], -1)
|
70 |
+
crop = masked[y : y + h, x : x + w]
|
71 |
if h > w:
|
72 |
top, bottom, left, right = 0, 0, (h - w) // 2, (h - w) // 2
|
73 |
else:
|
|
|
82 |
cv2.BORDER_CONSTANT,
|
83 |
value=(0, 0, 0),
|
84 |
)
|
|
|
85 |
crop = PIL.Image.fromarray(crop)
|
86 |
return crop
|
87 |
|
88 |
|
89 |
+
def get_texts(query: str) -> List[str]:
|
90 |
+
return [f"a picture of {query}", "a picture of background"]
|
91 |
+
|
92 |
+
|
93 |
def filter_masks(
|
94 |
image: np.ndarray,
|
95 |
masks: List[Dict[str, Any]],
|
|
|
98 |
query: str,
|
99 |
clip_threshold: float,
|
100 |
) -> List[Dict[str, Any]]:
|
|
|
101 |
filtered_masks: List[Dict[str, Any]] = []
|
102 |
|
103 |
+
for mask in sorted(masks, key=lambda mask: mask["area"])[-TOP_K_OBJ:]:
|
104 |
if (
|
105 |
mask["predicted_iou"] < predicted_iou_threshold
|
106 |
or mask["stability_score"] < stability_score_threshold
|
107 |
or image.shape[:2] != mask["segmentation"].shape[:2]
|
108 |
+
or query
|
109 |
+
and get_score(crop_image(image, mask), get_texts(query)) < clip_threshold
|
110 |
):
|
111 |
continue
|
|
|
|
|
112 |
|
113 |
+
filtered_masks.append(mask)
|
|
|
|
|
|
|
|
|
|
|
|
|
114 |
|
115 |
return filtered_masks
|
116 |
|
|
|
132 |
contours, _ = cv2.findContours(
|
133 |
np.uint8(mask["segmentation"]), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE
|
134 |
)
|
135 |
+
cv2.drawContours(image, contours, -1, (0, 0, 255), 2)
|
136 |
return image
|
137 |
|
138 |
|
|
|
144 |
query: str,
|
145 |
) -> PIL.ImageFile.ImageFile:
|
146 |
mask_generator = load_mask_generator()
|
147 |
+
image = cv2.imread(image_path, cv2.IMREAD_COLOR)
|
148 |
+
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
|
149 |
+
|
150 |
# reduce the size to save gpu memory
|
151 |
+
image = adjust_image_size(image)
|
152 |
masks = mask_generator.generate(image)
|
153 |
masks = filter_masks(
|
154 |
image,
|
|
|
159 |
clip_threshold,
|
160 |
)
|
161 |
image = draw_masks(image, masks)
|
|
|
162 |
image = PIL.Image.fromarray(image)
|
163 |
return image
|
164 |
|
|
|
179 |
[
|
180 |
0.9,
|
181 |
0.8,
|
182 |
+
0.99,
|
183 |
os.path.join(os.path.dirname(__file__), "examples/dog.jpg"),
|
184 |
+
"dog",
|
185 |
],
|
186 |
[
|
187 |
0.9,
|
188 |
0.8,
|
189 |
+
0.75,
|
190 |
os.path.join(os.path.dirname(__file__), "examples/city.jpg"),
|
191 |
"building",
|
192 |
],
|
193 |
[
|
194 |
0.9,
|
195 |
0.8,
|
196 |
+
0.998,
|
197 |
os.path.join(os.path.dirname(__file__), "examples/food.jpg"),
|
198 |
+
"strawberry",
|
199 |
],
|
200 |
[
|
201 |
0.9,
|
202 |
0.8,
|
203 |
+
0.75,
|
204 |
os.path.join(os.path.dirname(__file__), "examples/horse.jpg"),
|
205 |
"horse",
|
206 |
],
|
207 |
+
[
|
208 |
+
0.9,
|
209 |
+
0.8,
|
210 |
+
0.99,
|
211 |
+
os.path.join(os.path.dirname(__file__), "examples/bears.jpg"),
|
212 |
+
"bear",
|
213 |
+
],
|
214 |
+
[
|
215 |
+
0.9,
|
216 |
+
0.8,
|
217 |
+
0.99,
|
218 |
+
os.path.join(os.path.dirname(__file__), "examples/cats.jpg"),
|
219 |
+
"cat",
|
220 |
+
],
|
221 |
+
[
|
222 |
+
0.9,
|
223 |
+
0.8,
|
224 |
+
0.99,
|
225 |
+
os.path.join(os.path.dirname(__file__), "examples/fish.jpg"),
|
226 |
+
"fish",
|
227 |
+
],
|
228 |
],
|
229 |
)
|
230 |
|
examples/bears.jpg
ADDED
examples/cats.jpg
ADDED
examples/fish.jpg
ADDED