Update app.py

app.py

@@ -5,13 +5,19 @@ from transformers import AutoModelForImageSegmentation
 import torch
 from torchvision import transforms
 
-torch.set_float32_matmul_precision(["high", "highest"][0])
+# Set float32 matmul precision (used for performance tuning)
+torch.set_float32_matmul_precision("high")
 
+# Detect device (GPU if available, otherwise CPU)
+device = "cuda" if torch.cuda.is_available() else "cpu"
+
+# Load model and send to appropriate device
 birefnet = AutoModelForImageSegmentation.from_pretrained(
     "ZhengPeng7/BiRefNet", trust_remote_code=True
 )
-birefnet.to("cpu")
+birefnet.to(device)
 
+# Image transformation pipeline
 transform_image = transforms.Compose(
     [
         transforms.Resize((1024, 1024)),
@@ -20,18 +26,10 @@ transform_image = transforms.Compose(
     ]
 )
 
-def fn(image):
-    im = load_img(image, output_type="pil")
-    im = im.convert("RGB")
-    origin = im.copy()
-    processed_image = process(im)
-    return (processed_image, origin)
-
-@spaces.GPU
+# Background removal pipeline
 def process(image):
     image_size = image.size
-    input_images = transform_image(image).unsqueeze(0).to("cuda")
-    # Prediction
+    input_images = transform_image(image).unsqueeze(0).to(device)
     with torch.no_grad():
         preds = birefnet(input_images)[-1].sigmoid().cpu()
     pred = preds[0].squeeze()
@@ -40,6 +38,15 @@ def process(image):
     image.putalpha(mask)
     return image
 
+# Gradio interface function
+def fn(image):
+    im = load_img(image, output_type="pil")
+    im = im.convert("RGB")
+    origin = im.copy()
+    processed_image = process(im)
+    return (processed_image, origin)
+
+# Process uploaded file and return PNG with alpha channel
 def process_file(f):
     name_path = f.rsplit(".", 1)[0] + ".png"
     im = load_img(f, output_type="pil")
@@ -48,6 +55,7 @@ def process_file(f):
     transparent.save(name_path)
     return name_path
 
+# Gradio components
 slider1 = gr.ImageSlider(label="Processed Image", type="pil", format="png")
 slider2 = gr.ImageSlider(label="Processed Image from URL", type="pil", format="png")
 image_upload = gr.Image(label="Upload an image")
@@ -59,13 +67,15 @@ output_file = gr.File(label="Output PNG File")
 chameleon = load_img("butterfly.jpg", output_type="pil")
 url_example = "https://hips.hearstapps.com/hmg-prod/images/gettyimages-1229892983-square.jpg"
 
+# Gradio interface tabs
 tab1 = gr.Interface(fn, inputs=image_upload, outputs=slider1, examples=[chameleon], api_name="image")
 tab2 = gr.Interface(fn, inputs=url_input, outputs=slider2, examples=[url_example], api_name="text")
 tab3 = gr.Interface(process_file, inputs=image_file_upload, outputs=output_file, examples=["butterfly.jpg"], api_name="png")
 
+# Launch app
 demo = gr.TabbedInterface(
     [tab1, tab2, tab3], ["Image Upload", "URL Input", "File Output"], title="Background Removal Tool"
 )
 
 if __name__ == "__main__":
-    demo.launch(show_error=True)
+    demo.launch(show_error=True)