Update app.py

app.py

@@ -1,49 +1,25 @@
-import cv2
+import gradio as gr
 import numpy as np
-from PIL import Image, ImageFilter
-from rembg import new_session
-from skimage import restoration
+from PIL import Image
+from rembg import remove, new_session
 
-# Initialize ISNet (best free-tier detail model)
+# Initialize session (will auto-download model)
 session = new_session("isnet-general-use")
 
-def perfect_remove_bg(img):
+def remove_bg(image):
     try:
-        # Convert input
-        if isinstance(img, np.ndarray):
-            img = Image.fromarray(img)
-        
-        # Preserve original size
-        w, h = img.size
-        
-        # Step 1: Initial removal
-        result = remove(
-            img,
-            session=session,
-            alpha_matting=False,  # Disable auto-crop
-            post_process_mask=True
-        )
-        
-        # Step 2: Edge refinement (CPU-only)
-        mask = np.array(result.split()[-1])  # Extract alpha
-        mask = restoration.denoise_tv_chambolle(mask, weight=0.1)  # Smooth edges
-        mask = (mask * 255).astype(np.uint8)
-        
-        # Step 3: Apply refined mask
-        result.putalpha(Image.fromarray(mask).resize((w,h)))
-        
-        return result
-        
+        if isinstance(image, np.ndarray):
+            image = Image.fromarray(image)
+        return remove(image, session=session)
     except Exception as e:
         print(f"Error: {e}")
-        return img
+        return image
 
-# Gradio interface
 with gr.Blocks() as demo:
-    gr.Markdown("## ✨ Free Detail-Preserving BG Remover")
+    gr.Markdown("## 🆓 Free Background Remover")
     with gr.Row():
-        input_img = gr.Image(label="Input", type="pil", sources=["upload"])
+        input_img = gr.Image(label="Input", type="pil")
         output_img = gr.Image(label="Output", type="pil")
-    gr.Button("Process").click(perfect_remove_bg, inputs=input_img, outputs=output_img)
+    gr.Button("Remove BG").click(remove_bg, inputs=input_img, outputs=output_img)
 
-demo.launch(show_error=True)
+demo.launch()