推理阶段不需要做图片预处理，直接扔到模型里面就好了。

classifier/food_classifier_app.py

@@ -57,19 +57,15 @@ class FoodClassifierApp:
         try:
             model_path = settings.INFERENCE_BEST_MODEL_PATH
             if os.path.exists(model_path):
-                # 创建模型实例
-                self.model = create_food_cnn()
+                # 创建模型实例（不需要内部预处理，因为我们使用preprocess_image方法）
+                self.model = create_food_cnn(use_internal_preprocess=False)
                 # 加载模型权重
                 self.model.load_state_dict(torch.load(model_path, map_location=self.device))
                 self.model.to(self.device)
                 self.model.eval()  # 设置为评估模式
                 
-                # 定义图像预处理（与训练时相同）
-                self.transform = transforms.Compose([
-                    transforms.Resize((32, 32),interpolation=transforms.InterpolationMode.BILINEAR),
-                    transforms.ToTensor(),
-                    transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))
-                ])
+                # 不再需要定义transform，直接使用模型的preprocess_image方法
+                self.transform = None
                 
                 print("PyTorch模型加载成功")
             else:
@@ -502,7 +498,7 @@ class FoodClassifierApp:
             
             for i, img_info in enumerate(self.uploaded_images):
                 # 预处理图片
-                if self.model is not None and self.transform is not None:
+                if self.model is not None:
                     # 使用真实模型预测
                     prediction, confidence = self.predict_with_model(img_info['image'])
                     predicted_class = self.food_classes[prediction]
@@ -549,25 +545,9 @@ class FoodClassifierApp:
             image_rgb = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
             pil_image = Image.fromarray(image_rgb)
             
-            # 应用预处理
-            resize_transform = transforms.Resize((32,32),interpolation=transforms.InterpolationMode.BICUBIC)
-            resized_image = resize_transform(pil_image)
-            if isinstance(resized_image, Image.Image):
-                # 转换为tensor但不归一化
-                to_tensor = transforms.ToTensor()
-                resized_tensor = to_tensor(resized_image)
-                print(f"缩放后tensor形状: {resized_tensor.shape}")
-
-                # 打印前5个像素值（每个通道）
-                print("前5个像素值 (R, G, B):")
-                for i in range(min(5, resized_tensor.shape[1])):
-                    r_val = resized_tensor[0, 0, i].item() * 255  # Red通道 (转换回0-255范围)
-                    g_val = resized_tensor[1, 0, i].item() * 255  # Green通道
-                    b_val = resized_tensor[2, 0, i].item() * 255  # Blue通道
-                    print(f"  像素[0,{i}]: R={r_val:.2f}, G={g_val:.2f}, B={b_val:.2f}")
-
-
-            input_tensor = self.transform(pil_image).unsqueeze(0)  # 添加batch维度
+            # 不再需要手动预处理，模型会自动处理
+            # 使用模型的预处理方法
+            input_tensor = self.model.preprocess_image(pil_image)
             input_tensor = input_tensor.to(self.device)
             
             # 进行预测