Spaces:
Running
on
Zero
Running
on
Zero
- app.py +5 -2
- sam2_mask.py +195 -0
app.py
CHANGED
@@ -11,6 +11,7 @@ from gradio_image_prompter import ImagePrompter
|
|
11 |
from PIL import Image, ImageDraw
|
12 |
import numpy as np
|
13 |
from sam2.sam2_image_predictor import SAM2ImagePredictor
|
|
|
14 |
import subprocess
|
15 |
|
16 |
subprocess.run("rm -rf /data-nvme/zerogpu-offload/*", env={}, shell=True)
|
@@ -557,6 +558,8 @@ with gr.Blocks(css=css, fill_height=True) as demo:
|
|
557 |
use_as_input_button_outpaint = gr.Button("Use as Input Image", visible=False)
|
558 |
history_gallery = gr.Gallery(label="History", columns=6, object_fit="contain", interactive=False)
|
559 |
preview_image = gr.Image(label="Preview")
|
|
|
|
|
560 |
with gr.TabItem("SAM2 Mask"):
|
561 |
gr.Markdown("# Object Segmentation with SAM2")
|
562 |
gr.Markdown(
|
@@ -571,9 +574,9 @@ with gr.Blocks(css=css, fill_height=True) as demo:
|
|
571 |
upload_image_input = ImagePrompter(show_label=False)
|
572 |
with gr.Column():
|
573 |
image_output = gr.Image(label="Segmented Image", type="pil", height=400)
|
574 |
-
|
575 |
# Button to trigger the prediction
|
576 |
-
|
577 |
|
578 |
# Define the action triggered by the predict button
|
579 |
predict_button.click(
|
|
|
11 |
from PIL import Image, ImageDraw
|
12 |
import numpy as np
|
13 |
from sam2.sam2_image_predictor import SAM2ImagePredictor
|
14 |
+
from sam2_mask import create_sam2_tab
|
15 |
import subprocess
|
16 |
|
17 |
subprocess.run("rm -rf /data-nvme/zerogpu-offload/*", env={}, shell=True)
|
|
|
558 |
use_as_input_button_outpaint = gr.Button("Use as Input Image", visible=False)
|
559 |
history_gallery = gr.Gallery(label="History", columns=6, object_fit="contain", interactive=False)
|
560 |
preview_image = gr.Image(label="Preview")
|
561 |
+
with gr.TabItem("SAM2 Masking"):
|
562 |
+
input_image, points_map, output_result_mask = create_sam2_tab()
|
563 |
with gr.TabItem("SAM2 Mask"):
|
564 |
gr.Markdown("# Object Segmentation with SAM2")
|
565 |
gr.Markdown(
|
|
|
574 |
upload_image_input = ImagePrompter(show_label=False)
|
575 |
with gr.Column():
|
576 |
image_output = gr.Image(label="Segmented Image", type="pil", height=400)
|
577 |
+
with gr.Row():
|
578 |
# Button to trigger the prediction
|
579 |
+
predict_button = gr.Button("Predict Mask")
|
580 |
|
581 |
# Define the action triggered by the predict button
|
582 |
predict_button.click(
|
sam2_mask.py
CHANGED
@@ -0,0 +1,195 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import spaces
|
2 |
+
import gradio as gr
|
3 |
+
import os
|
4 |
+
os.environ["TORCH_CUDNN_SDPA_ENABLED"] = "1"
|
5 |
+
import torch
|
6 |
+
import numpy as np
|
7 |
+
import cv2
|
8 |
+
import matplotlib.pyplot as plt
|
9 |
+
from PIL import Image, ImageFilter
|
10 |
+
from sam2.build_sam import build_sam2
|
11 |
+
from sam2.sam2_image_predictor import SAM2ImagePredictor
|
12 |
+
from gradio_image_prompter import ImagePrompter
|
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 |
+
|
20 |
+
# Extract x, y coordinates from evt.index
|
21 |
+
x, y = evt.index
|
22 |
+
|
23 |
+
# Add the point as [x, y]
|
24 |
+
tracking_points.append([x, y])
|
25 |
+
print(f"TRACKING POINTS: {tracking_points}")
|
26 |
+
|
27 |
+
if point_type == "include":
|
28 |
+
trackings_input_label.append(1)
|
29 |
+
elif point_type == "exclude":
|
30 |
+
trackings_input_label.append(0)
|
31 |
+
print(f"TRACKING INPUT LABELS: {trackings_input_label}")
|
32 |
+
|
33 |
+
# Open the image and get its dimensions
|
34 |
+
transparent_background = Image.open(first_frame_path).convert('RGBA')
|
35 |
+
w, h = transparent_background.size
|
36 |
+
|
37 |
+
# Define the circle radius as a fraction of the smaller dimension
|
38 |
+
fraction = 0.02 # You can adjust this value as needed
|
39 |
+
radius = int(fraction * min(w, h))
|
40 |
+
|
41 |
+
# Create a transparent layer to draw on
|
42 |
+
transparent_layer = np.zeros((h, w, 4), dtype=np.uint8)
|
43 |
+
|
44 |
+
for index, track in enumerate(tracking_points):
|
45 |
+
if trackings_input_label[index] == 1:
|
46 |
+
cv2.circle(transparent_layer, tuple(track), radius, (0, 255, 0, 255), -1)
|
47 |
+
else:
|
48 |
+
cv2.circle(transparent_layer, tuple(track), radius, (255, 0, 0, 255), -1)
|
49 |
+
|
50 |
+
# Convert the transparent layer back to an image
|
51 |
+
transparent_layer = Image.fromarray(transparent_layer, 'RGBA')
|
52 |
+
selected_point_map = Image.alpha_composite(transparent_background, transparent_layer)
|
53 |
+
|
54 |
+
return tracking_points, trackings_input_label, selected_point_map
|
55 |
+
|
56 |
+
def show_mask(mask, ax, random_color=False, borders=True):
|
57 |
+
if random_color:
|
58 |
+
color = np.concatenate([np.random.random(3), np.array([0.6])], axis=0)
|
59 |
+
else:
|
60 |
+
color = np.array([30/255, 144/255, 255/255, 0.6])
|
61 |
+
h, w = mask.shape[-2:]
|
62 |
+
mask = mask.astype(np.uint8)
|
63 |
+
mask_image = mask.reshape(h, w, 1) * color.reshape(1, 1, -1)
|
64 |
+
if borders:
|
65 |
+
import cv2
|
66 |
+
contours, _ = cv2.findContours(mask, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
|
67 |
+
# Try to smooth contours
|
68 |
+
contours = [cv2.approxPolyDP(contour, epsilon=0.01, closed=True) for contour in contours]
|
69 |
+
mask_image = cv2.drawContours(mask_image, contours, -1, (1, 1, 1, 0.5), thickness=2)
|
70 |
+
ax.imshow(mask_image)
|
71 |
+
|
72 |
+
def show_points(coords, labels, ax, marker_size=375):
|
73 |
+
pos_points = coords[labels == 1]
|
74 |
+
neg_points = coords[labels == 0]
|
75 |
+
ax.scatter(pos_points[:, 0], pos_points[:, 1], color='green', marker='*', s=marker_size, edgecolor='white', linewidth=1.25)
|
76 |
+
ax.scatter(neg_points[:, 0], neg_points[:, 1], color='red', marker='*', s=marker_size, edgecolor='white', linewidth=1.25)
|
77 |
+
|
78 |
+
def show_box(box, ax):
|
79 |
+
x0, y0 = box[0], box[1]
|
80 |
+
w, h = box[2] - box[0], box[3] - box[1]
|
81 |
+
ax.add_patch(plt.Rectangle((x0, y0), w, h, edgecolor='green', facecolor=(0, 0, 0, 0), lw=2))
|
82 |
+
|
83 |
+
def show_masks(image, masks, scores, point_coords=None, box_coords=None, input_labels=None, borders=True):
|
84 |
+
combined_images = [] # List to store filenames of images with masks overlaid
|
85 |
+
mask_images = [] # List to store filenames of separate mask images
|
86 |
+
for i, (mask, score) in enumerate(zip(masks, scores)):
|
87 |
+
# ---- Original Image with Mask Overlaid ----
|
88 |
+
plt.figure(figsize=(10, 10))
|
89 |
+
plt.imshow(image)
|
90 |
+
show_mask(mask, plt.gca(), borders=borders) # Draw the mask with borders
|
91 |
+
if box_coords is not None:
|
92 |
+
show_box(box_coords, plt.gca())
|
93 |
+
if len(scores) > 1:
|
94 |
+
plt.title(f"Mask {i+1}, Score: {score:.3f}", fontsize=18)
|
95 |
+
plt.axis('off')
|
96 |
+
# Save the figure as a JPG file
|
97 |
+
combined_filename = f"combined_image_{i+1}.jpg"
|
98 |
+
plt.savefig(combined_filename, format='jpg', bbox_inches='tight')
|
99 |
+
combined_images.append(combined_filename)
|
100 |
+
plt.close() # Close the figure to free up memory
|
101 |
+
# ---- Separate Mask Image (White Mask on Black Background) ----
|
102 |
+
# Create a black image
|
103 |
+
mask_image = np.zeros_like(image, dtype=np.uint8)
|
104 |
+
# The mask is a binary array where the masked area is 1, else 0.
|
105 |
+
# Convert the mask to a white color in the mask_image
|
106 |
+
mask_layer = (mask > 0).astype(np.uint8) * 255
|
107 |
+
for c in range(3): # Assuming RGB, repeat mask for all channels
|
108 |
+
mask_image[:, :, c] = mask_layer
|
109 |
+
# Save the mask image
|
110 |
+
mask_filename = f"mask_image_{i+1}.png"
|
111 |
+
Image.fromarray(mask_image).save(mask_filename)
|
112 |
+
mask_images.append(mask_filename)
|
113 |
+
plt.close() # Close the figure to free up memory
|
114 |
+
return combined_images, mask_images
|
115 |
+
|
116 |
+
@spaces.GPU()
|
117 |
+
def sam_process(original_image, points, labels):
|
118 |
+
print(f"Points: {points}")
|
119 |
+
print(f"Labels: {labels}")
|
120 |
+
if not points or not labels:
|
121 |
+
print("No points or labels provided, returning None")
|
122 |
+
return None
|
123 |
+
|
124 |
+
# Convert image to numpy array for SAM2 processing
|
125 |
+
image = np.array(original_image)
|
126 |
+
predictor = SAM2ImagePredictor.from_pretrained("facebook/sam2.1-hiera-large")
|
127 |
+
predictor.set_image(image)
|
128 |
+
|
129 |
+
input_point = np.array(points)
|
130 |
+
input_label = np.array(labels)
|
131 |
+
|
132 |
+
try:
|
133 |
+
masks, scores, _ = predictor.predict(input_point, input_label, multimask_output=False)
|
134 |
+
except Exception as e:
|
135 |
+
print(f"Error during prediction: {e}")
|
136 |
+
return None
|
137 |
+
|
138 |
+
sorted_indices = np.argsort(scores)[::-1]
|
139 |
+
masks = masks[sorted_indices]
|
140 |
+
|
141 |
+
if masks and len(masks) > 0:
|
142 |
+
mask = masks[0] * 255
|
143 |
+
mask_image = Image.fromarray(mask.astype(np.uint8))
|
144 |
+
return mask_image
|
145 |
+
else:
|
146 |
+
print("No masks generated, returning None")
|
147 |
+
return None
|
148 |
+
|
149 |
+
def create_sam2_tab():
|
150 |
+
first_frame = gr.State() # Tracks original image
|
151 |
+
tracking_points = gr.State([])
|
152 |
+
trackings_input_label = gr.State([])
|
153 |
+
|
154 |
+
with gr.Column():
|
155 |
+
gr.Markdown("# SAM2 Image Predictor")
|
156 |
+
|
157 |
+
image_input = gr.State()
|
158 |
+
input_image = ImagePrompter(show_label=False)
|
159 |
+
points_map = gr.Image(label="Points Map", type="pil", interactive=True)
|
160 |
+
# image_input = gr.Image(type="pil", visible=False) # Original image
|
161 |
+
|
162 |
+
with gr.Row():
|
163 |
+
point_type = gr.Radio(["include", "exclude"], value="include", label="Point Type")
|
164 |
+
clear_button = gr.Button("Clear Points")
|
165 |
+
|
166 |
+
submit_button = gr.Button("Submit")
|
167 |
+
output_image = gr.Image("Segmented Output")
|
168 |
+
|
169 |
+
# Event handlers
|
170 |
+
points_map.upload(
|
171 |
+
lambda img: (img, img, [], []),
|
172 |
+
inputs=points_map,
|
173 |
+
outputs=[input_image, first_frame, tracking_points, trackings_input_label]
|
174 |
+
)
|
175 |
+
|
176 |
+
clear_button.click(
|
177 |
+
lambda img: ([], [], img),
|
178 |
+
inputs=first_frame,
|
179 |
+
outputs=[tracking_points, trackings_input_label, points_map]
|
180 |
+
)
|
181 |
+
|
182 |
+
points_map.select(
|
183 |
+
get_point,
|
184 |
+
inputs=[point_type, tracking_points, trackings_input_label, first_frame],
|
185 |
+
outputs=[tracking_points, trackings_input_label, points_map]
|
186 |
+
)
|
187 |
+
|
188 |
+
submit_button.click(
|
189 |
+
sam_process,
|
190 |
+
inputs=[input_image, tracking_points, trackings_input_label],
|
191 |
+
outputs=output_image
|
192 |
+
)
|
193 |
+
|
194 |
+
return input_image, points_map, output_image
|
195 |
+
|