Update app.py

app.py

@@ -6,87 +6,67 @@
   from PIL import Image
   import cv2
 
-  class DepthEstimationAPI:
-      def __init__(self):
-          self.device = "cpu"  # Hugging Face Spacesは無料版でCPUのみ
-          print(f"Using device: {self.device}")
-
-          model_name = "depth-anything/Depth-Anything-V2-Small-hf"
-          self.processor = AutoImageProcessor.from_pretrained(model_name)     
-          self.model =
-  AutoModelForDepthEstimation.from_pretrained(model_name)
-          self.model.to(self.device)
-          self.model.eval()
-          print("Model loaded successfully")
-
-      def predict(self, image_input):
-          """Process image and return depth map"""
-          if image_input is None:
-              return None, None
-
-          try:
-              # PILイメージに変換
-              if hasattr(image_input, 'convert'):
-                  image = image_input.convert('RGB')
-              else:
-                  image = Image.open(image_input).convert('RGB')
-
-              # サイズ調整（メモリ節約）
-              max_size = 256
-              if max(image.size) > max_size:
-                  ratio = max_size / max(image.size)
-                  new_size = tuple(int(dim * ratio) for dim in
-  image.size)
-                  image = image.resize(new_size,
-  Image.Resampling.LANCZOS)
-
-              # 深度推定
-              inputs = self.processor(images=image, return_tensors="pt")      
-
-              with torch.no_grad():
-                  outputs = self.model(**inputs)
-                  depth = outputs.predicted_depth.squeeze().cpu().numpy()     
-
-              # 深度マップ可視化
-              depth_normalized = ((depth - depth.min()) / (depth.max() -      
+  # デバイス設定
+  device = "cpu"
+  print(f"Using device: {device}")
+
+  # モデル読み込み
+  model_name = "depth-anything/Depth-Anything-V2-Small-hf"
+  processor = AutoImageProcessor.from_pretrained(model_name)
+  model = AutoModelForDepthEstimation.from_pretrained(model_name)
+  model.to(device)
+  model.eval()
+  print("Model loaded successfully")
+
+  def predict_depth(image):
+      """深度推定関数"""
+      if image is None:
+          return None, None
+
+      try:
+          # 画像処理
+          if hasattr(image, 'convert'):
+              image = image.convert('RGB')
+
+          # サイズ調整
+          max_size = 256
+          if max(image.size) > max_size:
+              ratio = max_size / max(image.size)
+              new_size = tuple(int(dim * ratio) for dim in image.size)        
+              image = image.resize(new_size, Image.Resampling.LANCZOS)        
+
+          # 深度推定
+          inputs = processor(images=image, return_tensors="pt")
+
+          with torch.no_grad():
+              outputs = model(**inputs)
+              depth = outputs.predicted_depth.squeeze().cpu().numpy()
+
+          # 可視化
+          depth_norm = ((depth - depth.min()) / (depth.max() -
   depth.min()) * 255).astype(np.uint8)
-              depth_colored = cv2.applyColorMap(depth_normalized,
+          depth_colored = cv2.applyColorMap(depth_norm,
   cv2.COLORMAP_VIRIDIS)
-              depth_colored = cv2.cvtColor(depth_colored,
-  cv2.COLOR_BGR2RGB)
-              depth_image = Image.fromarray(depth_colored)
-
-              return image, depth_image
-
-          except Exception as e:
-              print(f"Error in prediction: {e}")
-              return image_input, None
-
-  # APIインスタンス初期化
-  api = DepthEstimationAPI()
-
-  # Gradioインターフェース作成
-  with gr.Blocks(title="Depth Estimation API") as demo:
-      gr.Markdown("# 深度推定 API")
-      gr.Markdown("DepthAnything V2を使用したAI深度推定")
-
-      with gr.Row():
-          with gr.Column():
-              input_image = gr.Image(type="pil",
-  label="画像をアップロード")
-              submit_btn = gr.Button("深度マップ生成", variant="primary")     
-
-          with gr.Column():
-              output_original = gr.Image(type="pil", label="元画像")
-              output_depth = gr.Image(type="pil", label="深度マップ")
-
-      # ボタンクリックで処理実行
-      submit_btn.click(
-          fn=api.predict,
-          inputs=[input_image],
-          outputs=[output_original, output_depth]
-      )
+          depth_colored = cv2.cvtColor(depth_colored, cv2.COLOR_BGR2RGB)      
+          depth_image = Image.fromarray(depth_colored)
+
+          return image, depth_image
+
+      except Exception as e:
+          print(f"Error: {e}")
+          return image, None
+
+  # Gradioインターフェース
+  demo = gr.Interface(
+      fn=predict_depth,
+      inputs=gr.Image(type="pil"),
+      outputs=[
+          gr.Image(type="pil", label="Original"),
+          gr.Image(type="pil", label="Depth Map")
+      ],
+      title="Depth Estimation API",
+      description="DepthAnything V2による深度推定"
+  )
 
-  # Hugging Face Spaces用起動設定
   if __name__ == "__main__":
       demo.launch()