反卷积可视化：使用ResNet18预训练模型做推理

        import torch
        import torch.nn as nn
        import torchvision.models as models
        import matplotlib.pyplot as plt
        # 加载预训练的 ResNet18 模型
        resnet18 = models.resnet18(weights=models.ResNet18_Weights.DEFAULT)

        # 设置为评估模式
        resnet18.eval()
        # 定义反卷积层
        deconv_layer = nn.ConvTranspose2d(in_channels=64, out_channels=3, kernel_size=3, stride=2, padding=1, output_padding=1)
        # 创建一个随机输入张量（假设输入图像为 3x224x224）
        input_tensor = torch.randn(1, 3, 224, 224)


        # 获取 ResNet18 的第一个卷积层输出
        with torch.no_grad():
            features = resnet18.conv1(input_tensor)
            features = resnet18.bn1(features)
            features = resnet18.relu(features)
            features = resnet18.maxpool(features)

        # 使用反卷积层对特征图进行上采样
        deconv_output = deconv_layer(features)

        # 将反卷积结果转换为图像格式
        output_image = deconv_output.squeeze().permute(1, 2, 0).detach().numpy()

        # 可视化结果
        plt.imshow(output_image)
        plt.axis('off')
        plt.show()                                  
        

        
        import torch
        import torch.nn as nn
        import torchvision.models as models
        import torchvision.transforms as transforms
        from PIL import Image
        import matplotlib.pyplot as plt

        # 1. 加载预训练的 ResNet18 模型
        resnet18 = models.resnet18(weights=models.ResNet18_Weights.DEFAULT)
        resnet18.eval()

        # 2. 定义反卷积层
        deconv_layer = nn.ConvTranspose2d(in_channels=64, out_channels=3, kernel_size=3, stride=2, padding=1, output_padding=1)

        # 3. 读取并预处理图片
        def load_image(image_path):
            # 打开图片
            image = Image.open(image_path).convert('RGB')  # 确保图片是 RGB 格式
            # 定义预处理步骤
            preprocess = transforms.Compose([
                transforms.Resize((224, 224)),  # 调整图片大小为 224x224
                transforms.ToTensor(),           # 转换为 PyTorch 张量
                transforms.Normalize(            # 标准化
                    mean=[0.485, 0.456, 0.406], # ImageNet 均值
                    std=[0.229, 0.224, 0.225]    # ImageNet 标准差
                )
            ])
            # 对图片进行预处理
            input_tensor = preprocess(image)
            input_tensor = input_tensor.unsqueeze(0)  # 添加 batch 维度
            return input_tensor, image

        # 4. 加载图片
        image_path = 'SAM/pic/suqi.jpg'  # 替换为你的图片路径
        input_tensor, original_image = load_image(image_path)

        # 5. 前向传播并获取特征图
        with torch.no_grad():
            features = resnet18.conv1(input_tensor)
            features = resnet18.bn1(features)
            features = resnet18.relu(features)
            features = resnet18.maxpool(features)

        # 6. 使用反卷积层对特征图进行上采样
        deconv_output = deconv_layer(features)

        # 7. 将反卷积结果转换为图像格式
        def deprocess_image(tensor):
            # 移除 batch 维度并转换为 numpy 数组
            tensor = tensor.squeeze(0).permute(1, 2, 0).detach().numpy()
            # 反标准化
            tensor = tensor * [0.229, 0.224, 0.225] + [0.485, 0.456, 0.406]
            # 将值限制在 [0, 1] 范围内
            tensor = tensor.clip(0, 1)
            return tensor

        output_image = deprocess_image(deconv_output)

        # 8. 可视化结果
        plt.figure(figsize=(10, 5))

        # 显示原始图片
        plt.subplot(1, 2, 1)
        plt.imshow(original_image)
        plt.title('Original Image')
        plt.axis('off')

        # 显示反卷积结果
        plt.subplot(1, 2, 2)
        plt.imshow(output_image)
        plt.title('Deconvolution Output')
        plt.axis('off')
        plt.imsave("SAM/pic/suqi-deconv.png",output_image)

        plt.show()