Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -6,23 +6,22 @@ from PIL import Image
|
|
6 |
from torchvision import transforms
|
7 |
from transformers import AutoModelForImageSegmentation
|
8 |
from typing import Union, List
|
9 |
-
from loadimg import load_img
|
10 |
|
11 |
torch.set_float32_matmul_precision("high")
|
12 |
|
13 |
-
# Load
|
14 |
-
|
15 |
-
"
|
16 |
-
trust_remote_code=True
|
17 |
)
|
18 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
19 |
-
|
20 |
|
21 |
-
#
|
22 |
transform_image = transforms.Compose([
|
23 |
transforms.Resize((1024, 1024)),
|
24 |
transforms.ToTensor(),
|
25 |
-
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
|
26 |
])
|
27 |
|
28 |
@spaces.GPU
|
@@ -31,28 +30,12 @@ def process(image: Image.Image) -> Image.Image:
|
|
31 |
input_tensor = transform_image(image).unsqueeze(0).to(device)
|
32 |
|
33 |
with torch.no_grad():
|
34 |
-
preds =
|
35 |
-
|
36 |
-
# Handle list output - extract the tensor from the list
|
37 |
-
if isinstance(preds, list):
|
38 |
-
# Usually the mask is the last or first element
|
39 |
-
pred = preds[-1] if len(preds) > 0 else preds[0]
|
40 |
-
elif isinstance(preds, tuple):
|
41 |
-
pred = preds[0]
|
42 |
-
else:
|
43 |
-
pred = preds
|
44 |
-
|
45 |
-
# Now apply sigmoid to the tensor
|
46 |
-
mask = pred.sigmoid().cpu()
|
47 |
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
-
|
52 |
-
# Create binary mask with threshold
|
53 |
-
binary_mask = mask_pil.point(lambda p: 255 if p > 127 else 0)
|
54 |
|
55 |
-
# Apply mask with white background
|
56 |
white_bg = Image.new("RGB", image_size, (255, 255, 255))
|
57 |
result = Image.composite(image, white_bg, binary_mask)
|
58 |
return result
|
@@ -62,6 +45,7 @@ def handler(image=None, image_url=None, batch_urls=None) -> Union[str, List[str]
|
|
62 |
results = []
|
63 |
|
64 |
try:
|
|
|
65 |
if image is not None:
|
66 |
image = image.convert("RGB")
|
67 |
processed = process(image)
|
@@ -69,6 +53,7 @@ def handler(image=None, image_url=None, batch_urls=None) -> Union[str, List[str]
|
|
69 |
processed.save(filename)
|
70 |
return filename
|
71 |
|
|
|
72 |
if image_url:
|
73 |
im = load_img(image_url, output_type="pil").convert("RGB")
|
74 |
processed = process(im)
|
@@ -76,6 +61,7 @@ def handler(image=None, image_url=None, batch_urls=None) -> Union[str, List[str]
|
|
76 |
processed.save(filename)
|
77 |
return filename
|
78 |
|
|
|
79 |
if batch_urls:
|
80 |
urls = [u.strip() for u in batch_urls.split(",") if u.strip()]
|
81 |
for url in urls:
|
@@ -91,11 +77,10 @@ def handler(image=None, image_url=None, batch_urls=None) -> Union[str, List[str]
|
|
91 |
|
92 |
except Exception as e:
|
93 |
print("General error:", e)
|
94 |
-
import traceback
|
95 |
-
traceback.print_exc()
|
96 |
|
97 |
return None
|
98 |
|
|
|
99 |
demo = gr.Interface(
|
100 |
fn=handler,
|
101 |
inputs=[
|
@@ -104,9 +89,9 @@ demo = gr.Interface(
|
|
104 |
gr.Textbox(label="Comma-separated Image URLs (Batch)"),
|
105 |
],
|
106 |
outputs=gr.File(label="Output File(s)", file_count="multiple"),
|
107 |
-
title="Background Remover (
|
108 |
description="Upload an image, paste a URL, or send a batch of URLs to remove the background and replace it with white.",
|
109 |
)
|
110 |
|
111 |
if __name__ == "__main__":
|
112 |
-
demo.launch(show_error=True, mcp_server=True)
|
|
|
6 |
from torchvision import transforms
|
7 |
from transformers import AutoModelForImageSegmentation
|
8 |
from typing import Union, List
|
9 |
+
from loadimg import load_img # Your helper to load from URL or file
|
10 |
|
11 |
torch.set_float32_matmul_precision("high")
|
12 |
|
13 |
+
# Load BiRefNet model
|
14 |
+
birefnet = AutoModelForImageSegmentation.from_pretrained(
|
15 |
+
"ZhengPeng7/BiRefNet", trust_remote_code=True
|
|
|
16 |
)
|
17 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
18 |
+
birefnet.to(device)
|
19 |
|
20 |
+
# Image transformation
|
21 |
transform_image = transforms.Compose([
|
22 |
transforms.Resize((1024, 1024)),
|
23 |
transforms.ToTensor(),
|
24 |
+
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
|
25 |
])
|
26 |
|
27 |
@spaces.GPU
|
|
|
30 |
input_tensor = transform_image(image).unsqueeze(0).to(device)
|
31 |
|
32 |
with torch.no_grad():
|
33 |
+
preds = birefnet(input_tensor)[-1].sigmoid().cpu()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
34 |
|
35 |
+
pred = preds[0].squeeze()
|
36 |
+
mask = transforms.ToPILImage()(pred).resize(image_size).convert("L")
|
37 |
+
binary_mask = mask.point(lambda p: 255 if p > 127 else 0)
|
|
|
|
|
|
|
38 |
|
|
|
39 |
white_bg = Image.new("RGB", image_size, (255, 255, 255))
|
40 |
result = Image.composite(image, white_bg, binary_mask)
|
41 |
return result
|
|
|
45 |
results = []
|
46 |
|
47 |
try:
|
48 |
+
# Single image upload
|
49 |
if image is not None:
|
50 |
image = image.convert("RGB")
|
51 |
processed = process(image)
|
|
|
53 |
processed.save(filename)
|
54 |
return filename
|
55 |
|
56 |
+
# Single image from URL
|
57 |
if image_url:
|
58 |
im = load_img(image_url, output_type="pil").convert("RGB")
|
59 |
processed = process(im)
|
|
|
61 |
processed.save(filename)
|
62 |
return filename
|
63 |
|
64 |
+
# Batch of URLs
|
65 |
if batch_urls:
|
66 |
urls = [u.strip() for u in batch_urls.split(",") if u.strip()]
|
67 |
for url in urls:
|
|
|
77 |
|
78 |
except Exception as e:
|
79 |
print("General error:", e)
|
|
|
|
|
80 |
|
81 |
return None
|
82 |
|
83 |
+
# Interface
|
84 |
demo = gr.Interface(
|
85 |
fn=handler,
|
86 |
inputs=[
|
|
|
89 |
gr.Textbox(label="Comma-separated Image URLs (Batch)"),
|
90 |
],
|
91 |
outputs=gr.File(label="Output File(s)", file_count="multiple"),
|
92 |
+
title="Background Remover (White Fill)",
|
93 |
description="Upload an image, paste a URL, or send a batch of URLs to remove the background and replace it with white.",
|
94 |
)
|
95 |
|
96 |
if __name__ == "__main__":
|
97 |
+
demo.launch(show_error=True, mcp_server=True)
|