Update app.py
Browse files
app.py
CHANGED
@@ -8,43 +8,35 @@ from PIL import Image
|
|
8 |
from sam2.build_sam import build_sam2
|
9 |
from sam2.sam2_image_predictor import SAM2ImagePredictor
|
10 |
|
|
|
|
|
|
|
11 |
def preprocess_image(image):
|
12 |
return image, gr.State([]), gr.State([]), image
|
13 |
|
14 |
def get_point(point_type, tracking_points, trackings_input_label, first_frame_path, evt: gr.SelectData):
|
15 |
print(f"You selected {evt.value} at {evt.index} from {evt.target}")
|
16 |
-
|
17 |
tracking_points.value.append(evt.index)
|
18 |
print(f"TRACKING POINT: {tracking_points.value}")
|
19 |
-
|
20 |
if point_type == "include":
|
21 |
trackings_input_label.value.append(1)
|
22 |
elif point_type == "exclude":
|
23 |
trackings_input_label.value.append(0)
|
24 |
print(f"TRACKING INPUT LABEL: {trackings_input_label.value}")
|
25 |
-
|
26 |
transparent_background = Image.open(first_frame_path).convert('RGBA')
|
27 |
w, h = transparent_background.size
|
28 |
-
|
29 |
fraction = 0.02
|
30 |
radius = int(fraction * min(w, h))
|
31 |
-
|
32 |
transparent_layer = np.zeros((h, w, 4), dtype=np.uint8)
|
33 |
-
|
34 |
for index, track in enumerate(tracking_points.value):
|
35 |
if trackings_input_label.value[index] == 1:
|
36 |
cv2.circle(transparent_layer, track, radius, (0, 255, 0, 255), -1)
|
37 |
else:
|
38 |
cv2.circle(transparent_layer, track, radius, (255, 0, 0, 255), -1)
|
39 |
-
|
40 |
transparent_layer = Image.fromarray(transparent_layer, 'RGBA')
|
41 |
selected_point_map = Image.alpha_composite(transparent_background, transparent_layer)
|
42 |
-
|
43 |
return tracking_points, trackings_input_label, selected_point_map
|
44 |
|
45 |
-
# Remove all CUDA-specific configurations
|
46 |
-
torch.autocast(device_type="cpu", dtype=torch.float32).__enter__()
|
47 |
-
|
48 |
def show_mask(mask, ax, random_color=False, borders=True):
|
49 |
if random_color:
|
50 |
color = np.concatenate([np.random.random(3), np.array([0.6])], axis=0)
|
@@ -54,9 +46,9 @@ def show_mask(mask, ax, random_color=False, borders=True):
|
|
54 |
mask = mask.astype(np.uint8)
|
55 |
mask_image = mask.reshape(h, w, 1) * color.reshape(1, 1, -1)
|
56 |
if borders:
|
57 |
-
contours, _
|
58 |
contours = [cv2.approxPolyDP(contour, epsilon=0.01, closed=True) for contour in contours]
|
59 |
-
mask_image = cv2.drawContours(mask_image, contours, -1, (1, 1, 1, 0.5), thickness=2)
|
60 |
ax.imshow(mask_image)
|
61 |
|
62 |
def show_points(coords, labels, ax, marker_size=375):
|
@@ -73,65 +65,82 @@ def show_box(box, ax):
|
|
73 |
def show_masks(image, masks, scores, point_coords=None, box_coords=None, input_labels=None, borders=True):
|
74 |
combined_images = []
|
75 |
mask_images = []
|
76 |
-
|
77 |
for i, (mask, score) in enumerate(zip(masks, scores)):
|
78 |
plt.figure(figsize=(10, 10))
|
79 |
plt.imshow(image)
|
80 |
show_mask(mask, plt.gca(), borders=borders)
|
81 |
plt.axis('off')
|
82 |
-
|
83 |
combined_filename = f"combined_image_{i+1}.jpg"
|
84 |
plt.savefig(combined_filename, format='jpg', bbox_inches='tight')
|
85 |
combined_images.append(combined_filename)
|
86 |
plt.close()
|
87 |
-
|
88 |
mask_image = np.zeros_like(image, dtype=np.uint8)
|
89 |
mask_layer = (mask > 0).astype(np.uint8) * 255
|
90 |
for c in range(3):
|
91 |
mask_image[:, :, c] = mask_layer
|
92 |
-
|
93 |
mask_filename = f"mask_image_{i+1}.png"
|
94 |
Image.fromarray(mask_image).save(mask_filename)
|
95 |
mask_images.append(mask_filename)
|
96 |
-
|
97 |
return combined_images, mask_images
|
98 |
|
99 |
-
def
|
100 |
-
|
101 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
102 |
|
|
|
|
|
|
|
103 |
checkpoint_map = {
|
104 |
"tiny": ("./checkpoints/sam2_hiera_tiny.pt", "sam2_hiera_t.yaml"),
|
105 |
"small": ("./checkpoints/sam2_hiera_small.pt", "sam2_hiera_s.yaml"),
|
106 |
"base-plus": ("./checkpoints/sam2_hiera_base_plus.pt", "sam2_hiera_b+.yaml"),
|
107 |
"large": ("./checkpoints/sam2_hiera_large.pt", "sam2_hiera_l.yaml")
|
108 |
}
|
109 |
-
|
110 |
sam2_checkpoint, model_cfg = checkpoint_map[checkpoint]
|
111 |
-
|
112 |
# Use CPU for both model and computations
|
113 |
sam2_model = build_sam2(model_cfg, sam2_checkpoint, device="cpu")
|
114 |
predictor = SAM2ImagePredictor(sam2_model)
|
115 |
predictor.set_image(image)
|
116 |
-
|
117 |
input_point = np.array(tracking_points.value)
|
118 |
input_label = np.array(trackings_input_label.value)
|
119 |
-
|
120 |
masks, scores, logits = predictor.predict(
|
121 |
point_coords=input_point,
|
122 |
point_labels=input_label,
|
123 |
multimask_output=False,
|
124 |
)
|
125 |
-
|
126 |
sorted_ind = np.argsort(scores)[::-1]
|
127 |
masks = masks[sorted_ind]
|
128 |
scores = scores[sorted_ind]
|
129 |
-
|
130 |
-
|
131 |
-
|
132 |
-
|
|
|
|
|
|
|
133 |
borders=True)
|
134 |
-
|
135 |
return results[0], mask_results[0]
|
136 |
|
137 |
with gr.Blocks() as demo:
|
@@ -149,36 +158,43 @@ with gr.Blocks() as demo:
|
|
149 |
point_type = gr.Radio(label="point type", choices=["include", "exclude"], value="include")
|
150 |
clear_points_btn = gr.Button("Clear Points")
|
151 |
checkpoint = gr.Dropdown(label="Checkpoint", choices=["tiny", "small", "base-plus", "large"], value="base-plus")
|
|
|
|
|
|
|
|
|
|
|
|
|
152 |
submit_btn = gr.Button("Submit")
|
153 |
with gr.Column():
|
154 |
output_result = gr.Image()
|
155 |
output_result_mask = gr.Image()
|
156 |
-
|
157 |
clear_points_btn.click(
|
158 |
fn=preprocess_image,
|
159 |
inputs=input_image,
|
160 |
outputs=[first_frame_path, tracking_points, trackings_input_label, points_map],
|
161 |
queue=False
|
162 |
)
|
163 |
-
|
164 |
points_map.upload(
|
165 |
fn=preprocess_image,
|
166 |
inputs=[points_map],
|
167 |
outputs=[first_frame_path, tracking_points, trackings_input_label, input_image],
|
168 |
queue=False
|
169 |
)
|
170 |
-
|
171 |
points_map.select(
|
172 |
fn=get_point,
|
173 |
inputs=[point_type, tracking_points, trackings_input_label, first_frame_path],
|
174 |
outputs=[tracking_points, trackings_input_label, points_map],
|
175 |
queue=False
|
176 |
)
|
177 |
-
|
178 |
submit_btn.click(
|
179 |
fn=sam_process,
|
180 |
-
inputs=[input_image, checkpoint, tracking_points, trackings_input_label],
|
181 |
outputs=[output_result, output_result_mask]
|
182 |
)
|
|
|
|
|
|
|
|
|
|
|
183 |
|
184 |
demo.launch(show_api=False, show_error=True)
|
|
|
8 |
from sam2.build_sam import build_sam2
|
9 |
from sam2.sam2_image_predictor import SAM2ImagePredictor
|
10 |
|
11 |
+
# Remove all CUDA-specific configurations
|
12 |
+
torch.autocast(device_type="cpu", dtype=torch.float32).__enter__()
|
13 |
+
|
14 |
def preprocess_image(image):
|
15 |
return image, gr.State([]), gr.State([]), image
|
16 |
|
17 |
def get_point(point_type, tracking_points, trackings_input_label, first_frame_path, evt: gr.SelectData):
|
18 |
print(f"You selected {evt.value} at {evt.index} from {evt.target}")
|
|
|
19 |
tracking_points.value.append(evt.index)
|
20 |
print(f"TRACKING POINT: {tracking_points.value}")
|
|
|
21 |
if point_type == "include":
|
22 |
trackings_input_label.value.append(1)
|
23 |
elif point_type == "exclude":
|
24 |
trackings_input_label.value.append(0)
|
25 |
print(f"TRACKING INPUT LABEL: {trackings_input_label.value}")
|
|
|
26 |
transparent_background = Image.open(first_frame_path).convert('RGBA')
|
27 |
w, h = transparent_background.size
|
|
|
28 |
fraction = 0.02
|
29 |
radius = int(fraction * min(w, h))
|
|
|
30 |
transparent_layer = np.zeros((h, w, 4), dtype=np.uint8)
|
|
|
31 |
for index, track in enumerate(tracking_points.value):
|
32 |
if trackings_input_label.value[index] == 1:
|
33 |
cv2.circle(transparent_layer, track, radius, (0, 255, 0, 255), -1)
|
34 |
else:
|
35 |
cv2.circle(transparent_layer, track, radius, (255, 0, 0, 255), -1)
|
|
|
36 |
transparent_layer = Image.fromarray(transparent_layer, 'RGBA')
|
37 |
selected_point_map = Image.alpha_composite(transparent_background, transparent_layer)
|
|
|
38 |
return tracking_points, trackings_input_label, selected_point_map
|
39 |
|
|
|
|
|
|
|
40 |
def show_mask(mask, ax, random_color=False, borders=True):
|
41 |
if random_color:
|
42 |
color = np.concatenate([np.random.random(3), np.array([0.6])], axis=0)
|
|
|
46 |
mask = mask.astype(np.uint8)
|
47 |
mask_image = mask.reshape(h, w, 1) * color.reshape(1, 1, -1)
|
48 |
if borders:
|
49 |
+
contours, _= cv2.findContours(mask,cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
|
50 |
contours = [cv2.approxPolyDP(contour, epsilon=0.01, closed=True) for contour in contours]
|
51 |
+
mask_image = cv2.drawContours(mask_image, contours, -1, (1, 1, 1, 0.5), thickness=2)
|
52 |
ax.imshow(mask_image)
|
53 |
|
54 |
def show_points(coords, labels, ax, marker_size=375):
|
|
|
65 |
def show_masks(image, masks, scores, point_coords=None, box_coords=None, input_labels=None, borders=True):
|
66 |
combined_images = []
|
67 |
mask_images = []
|
|
|
68 |
for i, (mask, score) in enumerate(zip(masks, scores)):
|
69 |
plt.figure(figsize=(10, 10))
|
70 |
plt.imshow(image)
|
71 |
show_mask(mask, plt.gca(), borders=borders)
|
72 |
plt.axis('off')
|
|
|
73 |
combined_filename = f"combined_image_{i+1}.jpg"
|
74 |
plt.savefig(combined_filename, format='jpg', bbox_inches='tight')
|
75 |
combined_images.append(combined_filename)
|
76 |
plt.close()
|
|
|
77 |
mask_image = np.zeros_like(image, dtype=np.uint8)
|
78 |
mask_layer = (mask > 0).astype(np.uint8) * 255
|
79 |
for c in range(3):
|
80 |
mask_image[:, :, c] = mask_layer
|
|
|
81 |
mask_filename = f"mask_image_{i+1}.png"
|
82 |
Image.fromarray(mask_image).save(mask_filename)
|
83 |
mask_images.append(mask_filename)
|
|
|
84 |
return combined_images, mask_images
|
85 |
|
86 |
+
def expand_contract_mask(mask, px, expand=True):
|
87 |
+
kernel = np.ones((px, px), np.uint8)
|
88 |
+
if expand:
|
89 |
+
return cv2.dilate(mask, kernel, iterations=1)
|
90 |
+
else:
|
91 |
+
return cv2.erode(mask, kernel, iterations=1)
|
92 |
+
|
93 |
+
def feather_mask(mask, feather_size=10):
|
94 |
+
feathered_mask = mask.copy()
|
95 |
+
Feathered_region = mask > 0
|
96 |
+
Feathered_region = cv2.dilate(Feathered_region.astype(np.uint8), np.ones((feather_size, feather_size), np.uint8), iterations=1)
|
97 |
+
Feathered_region = Feathered_region & (~mask.astype(bool))
|
98 |
+
|
99 |
+
for i in range(1, feather_size + 1):
|
100 |
+
weight = i / (feather_size + 1)
|
101 |
+
feathered_mask[Feathered_region] = feathered_mask[Feathered_region] * (1 - weight) + weight
|
102 |
+
|
103 |
+
return feathered_mask
|
104 |
+
|
105 |
+
def process_mask(mask, expand_contract_px, expand, feathering_enabled, feather_size):
|
106 |
+
if expand_contract_px > 0:
|
107 |
+
mask = expand_contract_mask(mask, expand_contract_px, expand)
|
108 |
+
if feathering_enabled:
|
109 |
+
mask = feather_mask(mask, feather_size)
|
110 |
+
return mask
|
111 |
|
112 |
+
def sam_process(input_image, checkpoint, tracking_points, trackings_input_label, expand_contract_px, expand, feathering_enabled, feather_size):
|
113 |
+
image = Image.open(input_image)
|
114 |
+
image = np.array.array(image.convert("RGB"))
|
115 |
checkpoint_map = {
|
116 |
"tiny": ("./checkpoints/sam2_hiera_tiny.pt", "sam2_hiera_t.yaml"),
|
117 |
"small": ("./checkpoints/sam2_hiera_small.pt", "sam2_hiera_s.yaml"),
|
118 |
"base-plus": ("./checkpoints/sam2_hiera_base_plus.pt", "sam2_hiera_b+.yaml"),
|
119 |
"large": ("./checkpoints/sam2_hiera_large.pt", "sam2_hiera_l.yaml")
|
120 |
}
|
|
|
121 |
sam2_checkpoint, model_cfg = checkpoint_map[checkpoint]
|
|
|
122 |
# Use CPU for both model and computations
|
123 |
sam2_model = build_sam2(model_cfg, sam2_checkpoint, device="cpu")
|
124 |
predictor = SAM2ImagePredictor(sam2_model)
|
125 |
predictor.set_image(image)
|
|
|
126 |
input_point = np.array(tracking_points.value)
|
127 |
input_label = np.array(trackings_input_label.value)
|
|
|
128 |
masks, scores, logits = predictor.predict(
|
129 |
point_coords=input_point,
|
130 |
point_labels=input_label,
|
131 |
multimask_output=False,
|
132 |
)
|
|
|
133 |
sorted_ind = np.argsort(scores)[::-1]
|
134 |
masks = masks[sorted_ind]
|
135 |
scores = scores[sorted_ind]
|
136 |
+
processed_masks = []
|
137 |
+
for mask in masks:
|
138 |
+
processed_mask = process_mask(mask, expand_contract_px, expand, feathering_enabled, feather_size)
|
139 |
+
processed_masks.append(processed_mask)
|
140 |
+
results, mask_results = show_masks(image, processed_masks, scores,
|
141 |
+
point_coords=input_point,
|
142 |
+
input_labels=input_label,
|
143 |
borders=True)
|
|
|
144 |
return results[0], mask_results[0]
|
145 |
|
146 |
with gr.Blocks() as demo:
|
|
|
158 |
point_type = gr.Radio(label="point type", choices=["include", "exclude"], value="include")
|
159 |
clear_points_btn = gr.Button("Clear Points")
|
160 |
checkpoint = gr.Dropdown(label="Checkpoint", choices=["tiny", "small", "base-plus", "large"], value="base-plus")
|
161 |
+
with gr.Row():
|
162 |
+
expand_contract_px = gr.Slider(minimum=0, maximum=50, default=0, label="Expand/Contract (pixels)")
|
163 |
+
expand = gr.Radio(["Expand", "Contract"], default="Expand", label="Action")
|
164 |
+
with gr.Row():
|
165 |
+
feathering_enabled = gr.Checkbox(default=False, label="Enable Feathering")
|
166 |
+
feather_size = gr.Slider(minimum=1, maximum=50, default=10, label="Feathering Size", visible=False)
|
167 |
submit_btn = gr.Button("Submit")
|
168 |
with gr.Column():
|
169 |
output_result = gr.Image()
|
170 |
output_result_mask = gr.Image()
|
|
|
171 |
clear_points_btn.click(
|
172 |
fn=preprocess_image,
|
173 |
inputs=input_image,
|
174 |
outputs=[first_frame_path, tracking_points, trackings_input_label, points_map],
|
175 |
queue=False
|
176 |
)
|
|
|
177 |
points_map.upload(
|
178 |
fn=preprocess_image,
|
179 |
inputs=[points_map],
|
180 |
outputs=[first_frame_path, tracking_points, trackings_input_label, input_image],
|
181 |
queue=False
|
182 |
)
|
|
|
183 |
points_map.select(
|
184 |
fn=get_point,
|
185 |
inputs=[point_type, tracking_points, trackings_input_label, first_frame_path],
|
186 |
outputs=[tracking_points, trackings_input_label, points_map],
|
187 |
queue=False
|
188 |
)
|
|
|
189 |
submit_btn.click(
|
190 |
fn=sam_process,
|
191 |
+
inputs=[input_image, checkpoint, tracking_points, trackings_input_label, expand_contract_px, expand, feathering_enabled, feather_size],
|
192 |
outputs=[output_result, output_result_mask]
|
193 |
)
|
194 |
+
feathering_enabled.change(
|
195 |
+
fn=lambda enabled: gr.update(visible=enabled),
|
196 |
+
inputs=[feathering_enabled],
|
197 |
+
outputs=[feather_size]
|
198 |
+
)
|
199 |
|
200 |
demo.launch(show_api=False, show_error=True)
|