Update app.py

app.py

@@ -4,158 +4,76 @@ import torch.nn.functional as F
 from torchvision.transforms.functional import normalize
 from huggingface_hub import hf_hub_download
 import gradio as gr
-from gradio_imageslider import ImageSlider
 from briarmbg import BriaRMBG
-import PIL
 from PIL import Image
-from typing import Tuple
 
-net=BriaRMBG()
-# model_path = "./model1.pth"
+# 모델 초기화 및 로드
+net = BriaRMBG()
 model_path = hf_hub_download("briaai/RMBG-1.4", 'model.pth')
 if torch.cuda.is_available():
     net.load_state_dict(torch.load(model_path))
-    net=net.cuda()
+    net = net.cuda()
 else:
-    net.load_state_dict(torch.load(model_path,map_location="cpu"))
-net.eval() 
+    net.load_state_dict(torch.load(model_path, map_location="cpu"))
+net.eval()
 
-    
-def resize_image(image):
+def resize_image(image, model_input_size=(1024, 1024)):
     image = image.convert('RGB')
-    model_input_size = (1024, 1024)
     image = image.resize(model_input_size, Image.BILINEAR)
     return image
 
-
-def process(image):
-
-    # prepare input
-    orig_image = Image.fromarray(image)
-    w,h = orig_im_size = 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])
+def process(image, background_image=None):
+    # 이미지 전처리
+    orig_image = Image.fromarray(image.astype(np.uint8))
+    resized_image = resize_image(orig_image)
+    im_tensor = torch.tensor(np.array(resized_image), dtype=torch.float32).permute(2, 0, 1).unsqueeze(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()
-
-    #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')
-
-    return new_im
-    # return [new_orig_image, new_im]
-
-
-# block = gr.Blocks().queue()
-
-# with block:
-#     gr.Markdown("## BRIA RMBG 1.4")
-#     gr.HTML('''
-#       <p style="margin-bottom: 10px; font-size: 94%">
-#         This is a demo for BRIA RMBG 1.4 that using
-#         <a href="https://huggingface.co/briaai/RMBG-1.4" target="_blank">BRIA RMBG-1.4 image matting model</a> as backbone. 
-#       </p>
-#     ''')
-#     with gr.Row():
-#         with gr.Column():
-#             input_image = gr.Image(sources=None, type="pil") # None for upload, ctrl+v and webcam
-#             # input_image = gr.Image(sources=None, type="numpy") # None for upload, ctrl+v and webcam
-#             run_button = gr.Button(value="Run")
-            
-#         with gr.Column():
-#             result_gallery = gr.Gallery(label='Output', show_label=False, elem_id="gallery", columns=[1], height='auto')
-#     ips = [input_image]
-#     run_button.click(fn=process, inputs=ips, outputs=[result_gallery])
-
-# block.launch(debug = True)
-
-# block = gr.Blocks().queue()
-
-gr.Markdown("## BRIA RMBG 1.4")
-gr.HTML('''
-  <p style="margin-bottom: 10px; font-size: 94%">
-    This is a demo for BRIA RMBG 1.4 that using
-    <a href="https://huggingface.co/briaai/RMBG-1.4" target="_blank">BRIA RMBG-1.4 image matting model</a> as backbone. 
-  </p>
-''')
-title = "Background Removal"
-description = r"""Background removal model developed by <a href='https://BRIA.AI' target='_blank'><b>BRIA.AI</b></a>, trained on a carefully selected dataset and is available as an open-source model for non-commercial use.<br> 
-For test upload your image and wait. Read more at model card <a href='https://huggingface.co/briaai/RMBG-1.4' target='_blank'><b>briaai/RMBG-1.4</b></a>.<br>
-"""
-examples = [['./input.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)
+        im_tensor = im_tensor.cuda()
+    
+    # 배경 제거 모델 추론
+    with torch.no_grad():
+        output = net(im_tensor)
+    output = F.interpolate(output[0][0], size=orig_image.size[::-1], mode='bilinear', align_corners=False)
+    output = torch.sigmoid(output).cpu().numpy().squeeze()
+    mask = (output * 255).astype(np.uint8)
+    new_im = Image.fromarray(mask).convert("L")
+
+    result_image = Image.new("RGBA", orig_image.size)
+    orig_image_rgba = orig_image.convert("RGBA")
+    result_image.paste(orig_image_rgba, mask=new_im)
+
+    # 배경 이미지가 제공된 경우, 합성
+    if background_image is not None:
+        result_image = merge_images(background_image, result_image)
 
+    return result_image
 
 def merge_images(background_path, foreground_image):
-    """
-    배경 이미지에 배경이 제거된 이미지를 투명하게 삽입합니다.
-    배경이 제거된 이미지는 배경 이미지 중앙에 30% 크기로 축소되어 삽입됩니다.
-    """
-    # 배경 이미지 로드
     background = Image.open(background_path).convert("RGBA")
-    # 전경(배경이 제거된 이미지) 이미지 로드
     foreground = foreground_image.convert("RGBA")
     
-    # 전경 이미지를 배경 이미지의 30% 크기로 조정
     scale_factor = 0.3
     new_size = (int(background.width * scale_factor), int(foreground.height * background.width / foreground.width * scale_factor))
     foreground_resized = foreground.resize(new_size, Image.Resampling.LANCZOS)
     
-    # 전경 이미지를 배경 이미지의 가운데에 위치시키기 위한 좌표 계산
     x = (background.width - new_size[0]) // 2
     y = (background.height - new_size[1]) // 2
     
-    # 배경 이미지 위에 전경 이미지를 붙임
     background.paste(foreground_resized, (x, y), foreground_resized)
-    
     return background
 
-def process(image, background_image=None):
-    """
-    이미지에서 배경을 제거합니다. 선택적으로, 사용자가 배경 이미지를 제공한 경우
-    배경 이미지 위에 배경이 제거된 이미지를 투명하게 삽입합니다.
-    """
-    # 배경 제거 로직을 여기에 구현합니다...
-    # 예를 들어, 배경 제거 처리 후 이미지를 new_im에 할당합니다.
-    # 이 예제에서는 단순화를 위해 image를 직접 사용합니다.
-    new_im = image  # 실제 구현에서는 여기에 배경 제거 로직의 결과를 할당해야 합니다.
-
-    # 배경 이미지가 제공되었는지 확인
-    if background_image is not None:
-        # 배경 이미지에 배경이 제거된 이미지를 삽입하는 로직을 구현합니다.
-        result_image = merge_images(background_image, new_im)
-    else:
-        result_image = new_im
-
-    return result_image
-
-
 # Gradio 인터페이스 정의
 demo = gr.Interface(
     fn=process,
     inputs=[
-        gr.Image(type="pil", label="Image to remove background"),
-        gr.Image(type="pil", label="Optional: Background Image")  # optional 키워드 인자 제거
+        gr.Image(type="numpy", label="Image to remove background"),
+        gr.Image(type="pil", label="Optional: Background Image")
     ],
     outputs="image",
-    examples=examples,
-    title=title,
-    description=description
+    title="Background Removal",
+    description="This is a demo for BRIA RMBG 1.4 that uses the BRIA RMBG-1.4 image matting model as backbone."
 )
+
+if __name__ == "__main__":
+    demo.launch()