Spaces:
Runtime error
Runtime error
save image embeddings in gradio session to avoid repeatedly encoding
Browse files- app.py +10 -8
- segment_anything/onnx/predictor_onnx.py +7 -2
- segment_anything/predictor.py +14 -6
app.py
CHANGED
@@ -107,6 +107,7 @@ def reset(session_state):
|
|
107 |
session_state['box_list'] = []
|
108 |
session_state['ori_image'] = None
|
109 |
session_state['image_with_prompt'] = None
|
|
|
110 |
return None, session_state
|
111 |
|
112 |
|
@@ -116,6 +117,7 @@ def reset_all(session_state):
|
|
116 |
session_state['box_list'] = []
|
117 |
session_state['ori_image'] = None
|
118 |
session_state['image_with_prompt'] = None
|
|
|
119 |
return None, None, session_state
|
120 |
|
121 |
|
@@ -145,8 +147,8 @@ def on_image_upload(
|
|
145 |
session_state['ori_image'] = copy.deepcopy(image)
|
146 |
session_state['image_with_prompt'] = copy.deepcopy(image)
|
147 |
print("Image changed")
|
148 |
-
|
149 |
-
|
150 |
|
151 |
return image, session_state
|
152 |
|
@@ -188,13 +190,11 @@ def segment_with_points(
|
|
188 |
)
|
189 |
image = session_state['image_with_prompt']
|
190 |
|
191 |
-
nd_image = np.array(session_state['ori_image'])
|
192 |
-
predictor.set_image(nd_image)
|
193 |
-
|
194 |
if ENABLE_ONNX:
|
195 |
coord_np = np.array(session_state['coord_list'])[None]
|
196 |
label_np = np.array(session_state['label_list'])[None]
|
197 |
masks, scores, _ = predictor.predict(
|
|
|
198 |
point_coords=coord_np,
|
199 |
point_labels=label_np,
|
200 |
)
|
@@ -204,6 +204,7 @@ def segment_with_points(
|
|
204 |
coord_np = np.array(session_state['coord_list'])
|
205 |
label_np = np.array(session_state['label_list'])
|
206 |
masks, scores, logits = predictor.predict(
|
|
|
207 |
point_coords=coord_np,
|
208 |
point_labels=label_np,
|
209 |
num_multimask_outputs=4,
|
@@ -271,18 +272,18 @@ def segment_with_box(
|
|
271 |
)
|
272 |
|
273 |
box_np = np.array(box)
|
274 |
-
nd_image = np.array(session_state['ori_image'])
|
275 |
-
predictor.set_image(nd_image)
|
276 |
if ENABLE_ONNX:
|
277 |
point_coords = box_np.reshape(2, 2)[None]
|
278 |
point_labels = np.array([2, 3])[None]
|
279 |
masks, _, _ = predictor.predict(
|
|
|
280 |
point_coords=point_coords,
|
281 |
point_labels=point_labels,
|
282 |
)
|
283 |
annotations = masks[:, 0, :, :]
|
284 |
else:
|
285 |
masks, scores, _ = predictor.predict(
|
|
|
286 |
box=box_np,
|
287 |
num_multimask_outputs=1,
|
288 |
)
|
@@ -312,7 +313,8 @@ with gr.Blocks(css=css, title="EdgeSAM") as demo:
|
|
312 |
'label_list': [],
|
313 |
'box_list': [],
|
314 |
'ori_image': None,
|
315 |
-
'image_with_prompt': None
|
|
|
316 |
})
|
317 |
|
318 |
with gr.Row():
|
|
|
107 |
session_state['box_list'] = []
|
108 |
session_state['ori_image'] = None
|
109 |
session_state['image_with_prompt'] = None
|
110 |
+
session_state['feature'] = None
|
111 |
return None, session_state
|
112 |
|
113 |
|
|
|
117 |
session_state['box_list'] = []
|
118 |
session_state['ori_image'] = None
|
119 |
session_state['image_with_prompt'] = None
|
120 |
+
session_state['feature'] = None
|
121 |
return None, None, session_state
|
122 |
|
123 |
|
|
|
147 |
session_state['ori_image'] = copy.deepcopy(image)
|
148 |
session_state['image_with_prompt'] = copy.deepcopy(image)
|
149 |
print("Image changed")
|
150 |
+
nd_image = np.array(image)
|
151 |
+
session_state['feature'] = predictor.set_image(nd_image)
|
152 |
|
153 |
return image, session_state
|
154 |
|
|
|
190 |
)
|
191 |
image = session_state['image_with_prompt']
|
192 |
|
|
|
|
|
|
|
193 |
if ENABLE_ONNX:
|
194 |
coord_np = np.array(session_state['coord_list'])[None]
|
195 |
label_np = np.array(session_state['label_list'])[None]
|
196 |
masks, scores, _ = predictor.predict(
|
197 |
+
features=session_state['feature'],
|
198 |
point_coords=coord_np,
|
199 |
point_labels=label_np,
|
200 |
)
|
|
|
204 |
coord_np = np.array(session_state['coord_list'])
|
205 |
label_np = np.array(session_state['label_list'])
|
206 |
masks, scores, logits = predictor.predict(
|
207 |
+
features=session_state['feature'],
|
208 |
point_coords=coord_np,
|
209 |
point_labels=label_np,
|
210 |
num_multimask_outputs=4,
|
|
|
272 |
)
|
273 |
|
274 |
box_np = np.array(box)
|
|
|
|
|
275 |
if ENABLE_ONNX:
|
276 |
point_coords = box_np.reshape(2, 2)[None]
|
277 |
point_labels = np.array([2, 3])[None]
|
278 |
masks, _, _ = predictor.predict(
|
279 |
+
features=session_state['feature'],
|
280 |
point_coords=point_coords,
|
281 |
point_labels=point_labels,
|
282 |
)
|
283 |
annotations = masks[:, 0, :, :]
|
284 |
else:
|
285 |
masks, scores, _ = predictor.predict(
|
286 |
+
features=session_state['feature'],
|
287 |
box=box_np,
|
288 |
num_multimask_outputs=1,
|
289 |
)
|
|
|
313 |
'label_list': [],
|
314 |
'box_list': [],
|
315 |
'ori_image': None,
|
316 |
+
'image_with_prompt': None,
|
317 |
+
'feature': None
|
318 |
})
|
319 |
|
320 |
with gr.Row():
|
segment_anything/onnx/predictor_onnx.py
CHANGED
@@ -60,17 +60,22 @@ class SamPredictorONNX:
|
|
60 |
self.features = outputs[0]
|
61 |
self.is_image_set = True
|
62 |
|
|
|
|
|
63 |
def predict(
|
64 |
self,
|
|
|
65 |
point_coords: Optional[np.ndarray] = None,
|
66 |
point_labels: Optional[np.ndarray] = None,
|
67 |
) -> Tuple[np.ndarray, np.ndarray, np.ndarray]:
|
68 |
-
if not self.is_image_set:
|
69 |
raise RuntimeError("An image must be set with .set_image(...) before mask prediction.")
|
|
|
|
|
70 |
|
71 |
point_coords = self.transform.apply_coords(point_coords, self.original_size)
|
72 |
outputs = self.decoder.run(None, {
|
73 |
-
'image_embeddings':
|
74 |
'point_coords': point_coords.astype(np.float32),
|
75 |
'point_labels': point_labels.astype(np.float32)
|
76 |
})
|
|
|
60 |
self.features = outputs[0]
|
61 |
self.is_image_set = True
|
62 |
|
63 |
+
return self.features
|
64 |
+
|
65 |
def predict(
|
66 |
self,
|
67 |
+
features: np.ndarray = None,
|
68 |
point_coords: Optional[np.ndarray] = None,
|
69 |
point_labels: Optional[np.ndarray] = None,
|
70 |
) -> Tuple[np.ndarray, np.ndarray, np.ndarray]:
|
71 |
+
if features is None and not self.is_image_set:
|
72 |
raise RuntimeError("An image must be set with .set_image(...) before mask prediction.")
|
73 |
+
if features is None:
|
74 |
+
features = self.features
|
75 |
|
76 |
point_coords = self.transform.apply_coords(point_coords, self.original_size)
|
77 |
outputs = self.decoder.run(None, {
|
78 |
+
'image_embeddings': features,
|
79 |
'point_coords': point_coords.astype(np.float32),
|
80 |
'point_labels': point_labels.astype(np.float32)
|
81 |
})
|
segment_anything/predictor.py
CHANGED
@@ -37,7 +37,7 @@ class SamPredictor:
|
|
37 |
self,
|
38 |
image: np.ndarray,
|
39 |
image_format: str = "RGB",
|
40 |
-
) ->
|
41 |
"""
|
42 |
Calculates the image embeddings for the provided image, allowing
|
43 |
masks to be predicted with the 'predict' method.
|
@@ -59,14 +59,14 @@ class SamPredictor:
|
|
59 |
input_image_torch = torch.as_tensor(input_image, device=self.device)
|
60 |
input_image_torch = input_image_torch.permute(2, 0, 1).contiguous()[None, :, :, :]
|
61 |
|
62 |
-
self.set_torch_image(input_image_torch, image.shape[:2])
|
63 |
|
64 |
@torch.no_grad()
|
65 |
def set_torch_image(
|
66 |
self,
|
67 |
transformed_image: torch.Tensor,
|
68 |
original_image_size: Tuple[int, ...],
|
69 |
-
) ->
|
70 |
"""
|
71 |
Calculates the image embeddings for the provided image, allowing
|
72 |
masks to be predicted with the 'predict' method. Expects the input
|
@@ -91,8 +91,11 @@ class SamPredictor:
|
|
91 |
self.features = self.model.image_encoder(input_image)
|
92 |
self.is_image_set = True
|
93 |
|
|
|
|
|
94 |
def predict(
|
95 |
self,
|
|
|
96 |
point_coords: Optional[np.ndarray] = None,
|
97 |
point_labels: Optional[np.ndarray] = None,
|
98 |
box: Optional[np.ndarray] = None,
|
@@ -131,9 +134,12 @@ class SamPredictor:
|
|
131 |
of masks and H=W=256. These low resolution logits can be passed to
|
132 |
a subsequent iteration as mask input.
|
133 |
"""
|
134 |
-
if not self.is_image_set:
|
135 |
raise RuntimeError("An image must be set with .set_image(...) before mask prediction.")
|
136 |
|
|
|
|
|
|
|
137 |
# Transform input prompts
|
138 |
coords_torch, labels_torch, box_torch, mask_input_torch = None, None, None, None
|
139 |
if point_coords is not None:
|
@@ -153,6 +159,7 @@ class SamPredictor:
|
|
153 |
mask_input_torch = mask_input_torch[None, :, :, :]
|
154 |
|
155 |
masks, iou_predictions, low_res_masks = self.predict_torch(
|
|
|
156 |
coords_torch,
|
157 |
labels_torch,
|
158 |
box_torch,
|
@@ -170,6 +177,7 @@ class SamPredictor:
|
|
170 |
@torch.no_grad()
|
171 |
def predict_torch(
|
172 |
self,
|
|
|
173 |
point_coords: Optional[torch.Tensor],
|
174 |
point_labels: Optional[torch.Tensor],
|
175 |
boxes: Optional[torch.Tensor] = None,
|
@@ -211,7 +219,7 @@ class SamPredictor:
|
|
211 |
of masks and H=W=256. These low res logits can be passed to
|
212 |
a subsequent iteration as mask input.
|
213 |
"""
|
214 |
-
if not self.is_image_set:
|
215 |
raise RuntimeError("An image must be set with .set_image(...) before mask prediction.")
|
216 |
|
217 |
if point_coords is not None:
|
@@ -228,7 +236,7 @@ class SamPredictor:
|
|
228 |
|
229 |
# Predict masks
|
230 |
low_res_masks, iou_predictions = self.model.mask_decoder(
|
231 |
-
image_embeddings=
|
232 |
image_pe=self.model.prompt_encoder.get_dense_pe(),
|
233 |
sparse_prompt_embeddings=sparse_embeddings,
|
234 |
dense_prompt_embeddings=dense_embeddings,
|
|
|
37 |
self,
|
38 |
image: np.ndarray,
|
39 |
image_format: str = "RGB",
|
40 |
+
) -> torch.Tensor:
|
41 |
"""
|
42 |
Calculates the image embeddings for the provided image, allowing
|
43 |
masks to be predicted with the 'predict' method.
|
|
|
59 |
input_image_torch = torch.as_tensor(input_image, device=self.device)
|
60 |
input_image_torch = input_image_torch.permute(2, 0, 1).contiguous()[None, :, :, :]
|
61 |
|
62 |
+
return self.set_torch_image(input_image_torch, image.shape[:2])
|
63 |
|
64 |
@torch.no_grad()
|
65 |
def set_torch_image(
|
66 |
self,
|
67 |
transformed_image: torch.Tensor,
|
68 |
original_image_size: Tuple[int, ...],
|
69 |
+
) -> torch.Tensor:
|
70 |
"""
|
71 |
Calculates the image embeddings for the provided image, allowing
|
72 |
masks to be predicted with the 'predict' method. Expects the input
|
|
|
91 |
self.features = self.model.image_encoder(input_image)
|
92 |
self.is_image_set = True
|
93 |
|
94 |
+
return self.features
|
95 |
+
|
96 |
def predict(
|
97 |
self,
|
98 |
+
features: torch.Tensor = None,
|
99 |
point_coords: Optional[np.ndarray] = None,
|
100 |
point_labels: Optional[np.ndarray] = None,
|
101 |
box: Optional[np.ndarray] = None,
|
|
|
134 |
of masks and H=W=256. These low resolution logits can be passed to
|
135 |
a subsequent iteration as mask input.
|
136 |
"""
|
137 |
+
if features is None and not self.is_image_set:
|
138 |
raise RuntimeError("An image must be set with .set_image(...) before mask prediction.")
|
139 |
|
140 |
+
if features is None:
|
141 |
+
features = self.features
|
142 |
+
|
143 |
# Transform input prompts
|
144 |
coords_torch, labels_torch, box_torch, mask_input_torch = None, None, None, None
|
145 |
if point_coords is not None:
|
|
|
159 |
mask_input_torch = mask_input_torch[None, :, :, :]
|
160 |
|
161 |
masks, iou_predictions, low_res_masks = self.predict_torch(
|
162 |
+
features,
|
163 |
coords_torch,
|
164 |
labels_torch,
|
165 |
box_torch,
|
|
|
177 |
@torch.no_grad()
|
178 |
def predict_torch(
|
179 |
self,
|
180 |
+
features: torch.Tensor,
|
181 |
point_coords: Optional[torch.Tensor],
|
182 |
point_labels: Optional[torch.Tensor],
|
183 |
boxes: Optional[torch.Tensor] = None,
|
|
|
219 |
of masks and H=W=256. These low res logits can be passed to
|
220 |
a subsequent iteration as mask input.
|
221 |
"""
|
222 |
+
if features is None and not self.is_image_set:
|
223 |
raise RuntimeError("An image must be set with .set_image(...) before mask prediction.")
|
224 |
|
225 |
if point_coords is not None:
|
|
|
236 |
|
237 |
# Predict masks
|
238 |
low_res_masks, iou_predictions = self.model.mask_decoder(
|
239 |
+
image_embeddings=features,
|
240 |
image_pe=self.model.prompt_encoder.get_dense_pe(),
|
241 |
sparse_prompt_embeddings=sparse_embeddings,
|
242 |
dense_prompt_embeddings=dense_embeddings,
|