错误：给定组=1，权重大小为[32,1,3,3,3]，预期输入[1,4,193,229,193]有1个通道，但得到了4个通道

Question

RuntimeError: Given groups=1, weight of size [32, 1, 3, 3, 3], expected input[1, 4, 193, 229, 193]

有 1 个通道，但改为 4 个通道。

致力于 3D 卷积神经网络对帕金森病的分类。

我认为这部分代码有一些错误：

# Load the dataset using DataLoader
batch_size = 4

# Load the dataset with batch loading (adjust path as needed)
dataset = CustomDataset(root_dir=r'D:\PD\PD25_16bit', transform=ToTensor())
train_size = int(0.8 * len(dataset))
test_size = len(dataset) - train_size
train_dataset, test_dataset = torch.utils.data.random_split(dataset, [train_size, test_size])
train_loader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True)
test_loader = DataLoader(test_dataset, batch_size=batch_size, shuffle=False)

# Define the 3D CNN model
class CNN3D(nn.Module):
    def __init__(self, num_channels=1):
        super(CNN3D, self).__init__()
        self.conv1 = nn.Conv3d(num_channels, 32, kernel_size=3, stride=1, padding=1)
        self.pool = nn.MaxPool3d(kernel_size=2, stride=2)
        self.conv2 = nn.Conv3d(32, 64, kernel_size=3, stride=1, padding=1)
        self.fc1 = nn.Linear(64 * 48 * 57 * 48 // 4, 128)  # Adjust the linear layer input size
        self.fc2 = nn.Linear(128, 3)

    def forward(self, x):
        x = self.pool(F.relu(self.conv1(x)))
        x = self.pool(F.relu(self.conv2(x)))
        x = x.view(-1, (64 * 48 * 57 * 48 // 4))  # Adjust the view to handle batch size of 4
        x = F.relu(self.fc1(x))
        x = self.fc2(x)
        return x

每个3D文件（NIFTI）的尺寸为193,229,193。当我更改批量大小时，模型将其视为通道数。

Answer 1

Conv3d

函数需要形状为

(batch_size, channels, D, H, W)

的 5 维输入。您的输入具有单个通道，因此您需要添加单位维度。您的输入应该具有形状

(batch_size, 1, 193, 229, 193)