Spaces:
Starting
on
T4
Starting
on
T4
Update app.py
Browse files
app.py
CHANGED
@@ -29,9 +29,9 @@ def get_point(point_type, tracking_points, trackings_input_label, first_frame_pa
|
|
29 |
transparent_layer = np.zeros((h, w, 4))
|
30 |
for index, track in enumerate(tracking_points.value):
|
31 |
if trackings_input_label.value[index] == 1:
|
32 |
-
cv2.circle(transparent_layer, track,
|
33 |
else:
|
34 |
-
cv2.circle(transparent_layer, track,
|
35 |
|
36 |
transparent_layer = Image.fromarray(transparent_layer.astype(np.uint8))
|
37 |
selected_point_map = Image.alpha_composite(transparent_background, transparent_layer)
|
@@ -73,37 +73,53 @@ def show_box(box, ax):
|
|
73 |
w, h = box[2] - box[0], box[3] - box[1]
|
74 |
ax.add_patch(plt.Rectangle((x0, y0), w, h, edgecolor='green', facecolor=(0, 0, 0, 0), lw=2))
|
75 |
|
76 |
-
def show_masks(image, masks, scores, point_coords=None, box_coords=None, input_labels=None, borders=
|
77 |
-
|
|
|
|
|
78 |
for i, (mask, score) in enumerate(zip(masks, scores)):
|
|
|
79 |
plt.figure(figsize=(10, 10))
|
80 |
plt.imshow(image)
|
81 |
-
show_mask(mask, plt.gca(), borders=borders)
|
82 |
-
|
83 |
-
"""
|
84 |
if point_coords is not None:
|
85 |
assert input_labels is not None
|
86 |
show_points(point_coords, input_labels, plt.gca())
|
87 |
-
"""
|
88 |
-
|
89 |
if box_coords is not None:
|
90 |
-
# boxes
|
91 |
show_box(box_coords, plt.gca())
|
92 |
if len(scores) > 1:
|
93 |
plt.title(f"Mask {i+1}, Score: {score:.3f}", fontsize=18)
|
94 |
plt.axis('off')
|
95 |
-
# plt.show()
|
96 |
|
97 |
# Save the figure as a JPG file
|
98 |
-
|
99 |
-
plt.savefig(
|
|
|
|
|
|
|
100 |
|
101 |
-
|
102 |
-
|
103 |
-
|
104 |
-
plt.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
105 |
|
106 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
107 |
|
108 |
def sam_process(input_image, tracking_points, trackings_input_label):
|
109 |
image = Image.open(input_image)
|
@@ -135,10 +151,10 @@ def sam_process(input_image, tracking_points, trackings_input_label):
|
|
135 |
|
136 |
print(masks.shape)
|
137 |
|
138 |
-
results = show_masks(image, masks, scores, point_coords=input_point, input_labels=input_label, borders=
|
139 |
print(results)
|
140 |
|
141 |
-
return results[0]
|
142 |
|
143 |
with gr.Blocks() as demo:
|
144 |
first_frame_path = gr.State()
|
@@ -155,20 +171,33 @@ with gr.Blocks() as demo:
|
|
155 |
points_map = gr.Image(label="points map", interactive=False)
|
156 |
submit_btn = gr.Button("Submit")
|
157 |
output_result = gr.Image()
|
|
|
158 |
|
159 |
clear_points_btn.click(
|
160 |
fn = preprocess_image,
|
161 |
inputs = input_image,
|
162 |
-
outputs = [first_frame_path, tracking_points, trackings_input_label, points_map]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
163 |
)
|
164 |
-
input_image.upload(preprocess_image, input_image, [first_frame_path, tracking_points, trackings_input_label, points_map])
|
165 |
|
166 |
-
points_map.select(
|
|
|
|
|
|
|
|
|
|
|
167 |
|
168 |
|
169 |
submit_btn.click(
|
170 |
fn = sam_process,
|
171 |
inputs = [input_image, tracking_points, trackings_input_label],
|
172 |
-
outputs = [output_result]
|
173 |
)
|
174 |
demo.launch()
|
|
|
29 |
transparent_layer = np.zeros((h, w, 4))
|
30 |
for index, track in enumerate(tracking_points.value):
|
31 |
if trackings_input_label.value[index] == 1:
|
32 |
+
cv2.circle(transparent_layer, track, 20, (0, 0, 255, 255), -1)
|
33 |
else:
|
34 |
+
cv2.circle(transparent_layer, track, 20, (255, 0, 0, 255), -1)
|
35 |
|
36 |
transparent_layer = Image.fromarray(transparent_layer.astype(np.uint8))
|
37 |
selected_point_map = Image.alpha_composite(transparent_background, transparent_layer)
|
|
|
73 |
w, h = box[2] - box[0], box[3] - box[1]
|
74 |
ax.add_patch(plt.Rectangle((x0, y0), w, h, edgecolor='green', facecolor=(0, 0, 0, 0), lw=2))
|
75 |
|
76 |
+
def show_masks(image, masks, scores, point_coords=None, box_coords=None, input_labels=None, borders=True):
|
77 |
+
combined_images = [] # List to store filenames of images with masks overlaid
|
78 |
+
mask_images = [] # List to store filenames of separate mask images
|
79 |
+
|
80 |
for i, (mask, score) in enumerate(zip(masks, scores)):
|
81 |
+
# ---- Original Image with Mask Overlaid ----
|
82 |
plt.figure(figsize=(10, 10))
|
83 |
plt.imshow(image)
|
84 |
+
show_mask(mask, plt.gca(), borders=borders) # Draw the mask with borders
|
|
|
|
|
85 |
if point_coords is not None:
|
86 |
assert input_labels is not None
|
87 |
show_points(point_coords, input_labels, plt.gca())
|
|
|
|
|
88 |
if box_coords is not None:
|
|
|
89 |
show_box(box_coords, plt.gca())
|
90 |
if len(scores) > 1:
|
91 |
plt.title(f"Mask {i+1}, Score: {score:.3f}", fontsize=18)
|
92 |
plt.axis('off')
|
|
|
93 |
|
94 |
# Save the figure as a JPG file
|
95 |
+
combined_filename = f"combined_image_{i+1}.jpg"
|
96 |
+
plt.savefig(combined_filename, format='jpg', bbox_inches='tight')
|
97 |
+
combined_images.append(combined_filename)
|
98 |
+
|
99 |
+
plt.close() # Close the figure to free up memory
|
100 |
|
101 |
+
# ---- Separate Mask Image ----
|
102 |
+
plt.figure(figsize=(10, 10))
|
103 |
+
mask_image = np.zeros_like(image, dtype=np.uint8) # Initialize a blank image
|
104 |
+
show_mask(mask, plt.gca(), borders=False) # Draw the mask without borders
|
105 |
+
|
106 |
+
plt.axis('off')
|
107 |
+
plt.tight_layout()
|
108 |
+
plt.gca().set_axis_off()
|
109 |
+
plt.subplots_adjust(top=1, bottom=0, right=1, left=0,
|
110 |
+
hspace=0, wspace=0)
|
111 |
+
plt.margins(0, 0)
|
112 |
+
plt.gca().xaxis.set_major_locator(plt.NullLocator())
|
113 |
+
plt.gca().yaxis.set_major_locator(plt.NullLocator())
|
114 |
|
115 |
+
# Save mask image
|
116 |
+
mask_filename = f"mask_image_{i+1}.png"
|
117 |
+
plt.savefig(mask_filename, format='png', bbox_inches='tight', pad_inches=0)
|
118 |
+
mask_images.append(mask_filename)
|
119 |
+
|
120 |
+
plt.close() # Close the figure to free up memory
|
121 |
+
|
122 |
+
return combined_images, mask_images
|
123 |
|
124 |
def sam_process(input_image, tracking_points, trackings_input_label):
|
125 |
image = Image.open(input_image)
|
|
|
151 |
|
152 |
print(masks.shape)
|
153 |
|
154 |
+
results, mask_results = show_masks(image, masks, scores, point_coords=input_point, input_labels=input_label, borders=False)
|
155 |
print(results)
|
156 |
|
157 |
+
return results[0], mask_results[0]
|
158 |
|
159 |
with gr.Blocks() as demo:
|
160 |
first_frame_path = gr.State()
|
|
|
171 |
points_map = gr.Image(label="points map", interactive=False)
|
172 |
submit_btn = gr.Button("Submit")
|
173 |
output_result = gr.Image()
|
174 |
+
output_result_mask = gr.Image()
|
175 |
|
176 |
clear_points_btn.click(
|
177 |
fn = preprocess_image,
|
178 |
inputs = input_image,
|
179 |
+
outputs = [first_frame_path, tracking_points, trackings_input_label, points_map],
|
180 |
+
queue=False
|
181 |
+
)
|
182 |
+
|
183 |
+
input_image.upload(
|
184 |
+
preprocess_image,
|
185 |
+
input_image,
|
186 |
+
[first_frame_path, tracking_points, trackings_input_label, points_map],
|
187 |
+
queue=False
|
188 |
)
|
|
|
189 |
|
190 |
+
points_map.select(
|
191 |
+
get_point,
|
192 |
+
[point_type, tracking_points, trackings_input_label, first_frame_path],
|
193 |
+
[tracking_points, trackings_input_label, points_map],
|
194 |
+
queue=False
|
195 |
+
)
|
196 |
|
197 |
|
198 |
submit_btn.click(
|
199 |
fn = sam_process,
|
200 |
inputs = [input_image, tracking_points, trackings_input_label],
|
201 |
+
outputs = [output_result, output_result_mask]
|
202 |
)
|
203 |
demo.launch()
|