Update app.py

app.py

@@ -21,44 +21,41 @@ else:
 net.eval() 
 
     
-def resize_image(image):
+def resize_image(image, size=(1024, 1024)):
     image = image.convert('RGB')
-    model_input_size = (1024, 1024)
-    image = image.resize(model_input_size, Image.BILINEAR)
-    return image
+    return image.resize(size, Image.BILINEAR)
 
 
-def process(image):
-
-    # prepare input
-    orig_image = Image.fromarray(image)
-    w,h = orig_im_size = orig_image.size
+def process(image, background_image):
+    # Prepare and resize images
+    orig_image = Image.fromarray(image).convert("RGBA")
+    background_image = resize_image(Image.fromarray(background_image).convert("RGBA"))
+    
+    w, h = orig_image.size
     image = resize_image(orig_image)
     im_np = np.array(image)
-    im_tensor = torch.tensor(im_np, dtype=torch.float32).permute(2,0,1)
-    im_tensor = torch.unsqueeze(im_tensor,0)
-    im_tensor = torch.divide(im_tensor,255.0)
-    im_tensor = normalize(im_tensor,[0.5,0.5,0.5],[1.0,1.0,1.0])
+    im_tensor = torch.tensor(im_np, dtype=torch.float32).permute(2, 0, 1)
+    im_tensor = torch.unsqueeze(im_tensor, 0) / 255.0
+    im_tensor = normalize(im_tensor, [0.5, 0.5, 0.5], [1.0, 1.0, 1.0])
+    
     if torch.cuda.is_available():
-        im_tensor=im_tensor.cuda()
+        im_tensor = im_tensor.cuda()
 
-    #inference
-    result=net(im_tensor)
-    # post process
-    result = torch.squeeze(F.interpolate(result[0][0], size=(h,w), mode='bilinear') ,0)
-    ma = torch.max(result)
-    mi = torch.min(result)
-    result = (result-mi)/(ma-mi)    
-    # image to pil
-    im_array = (result*255).cpu().data.numpy().astype(np.uint8)
-    pil_im = Image.fromarray(np.squeeze(im_array))
-    # paste the mask on the original image
-    new_im = Image.new("RGBA", pil_im.size, (0,0,0,0))
-    new_im.paste(orig_image, mask=pil_im)
-    # new_orig_image = orig_image.convert('RGBA')
+    # Inference
+    result = net(im_tensor)
+    
+    # Post-process mask
+    result = torch.squeeze(F.interpolate(result[0][0], size=(h, w), mode='bilinear'), 0)
+    result = (result - result.min()) / (result.max() - result.min())
+    im_array = (result * 255).cpu().numpy().astype(np.uint8)
+    mask = Image.fromarray(im_array).convert("L")
 
-    return new_im
-    # return [new_orig_image, new_im]]
+    # Composite with new background
+    foreground = orig_image.copy()
+    new_background = background_image.resize((w, h))
+    new_background.paste(foreground, mask=mask)
+    
+    return new_background
     
 # block = gr.Blocks().queue()
 
@@ -96,10 +93,18 @@ title = "Human Body Segmentation"
 description = r"""Human Body Segmentation model developed by <a href='https://github.com/WildanJR09' target='_blank'><b>WildanJR</b></a>, Designed to effectively separate foreground from background in a range of categories and image types.<br> 
 This model has been trained on a carefully selected dataset, which includes: general stock images, e-commerce, gaming, and advertising content, making it suitable for commercial use cases powering enterprise content creation at scale. The accuracy, efficiency, and versatility currently rival leading source-available models. It is ideal where content safety, legally licensed datasets, and bias mitigation are paramount. For test upload your image and wait. </a>.<br>
 """
-examples = [['./jisoo.jpg'],]
+examples = [['./jisoo.jpg', './background.jpg']]
+
 # output = ImageSlider(position=0.5,label='Image without background', type="pil", show_download_button=True)
 # demo = gr.Interface(fn=process,inputs="image", outputs=output, examples=examples, title=title, description=description)
-demo = gr.Interface(fn=process,inputs="image", outputs="image", examples=examples, title=title, description=description)
+demo = gr.Interface(
+    fn=process,
+    inputs=["image", "image"],
+    outputs="image",
+    examples=examples,
+    title=title,
+    description=description
+)
 
 if __name__ == "__main__":
     demo.launch(share=False)