全连接层的输入，兼容所有图像尺寸

Question

    # Pooling layer
    self.pool = nn.MaxPool2d(kernel_size=2, stride=2)

    # fully connected layer
    self.fc1 = nn.Linear(256*59*59, 84)


    # Dropout layer
    self.dropout = nn.Dropout(p=0.5)

def forward(self, x):

    x = F.relu(self.conv1(x))
    x = F.relu(self.bn1(x))
    x = F.relu(self.conv2(x))
    x = F.relu(self.bn2(x))
    x = F.relu(self.conv3(x))
    x = F.relu(self.bn3(x))
    x = F.relu(self.conv4(x))
    x = F.relu(self.bn4(x))
    x = F.relu(self.conv5(x))
    x = F.relu(self.bn5(x))
    x = F.relu(self.pool(x))

    x = x.view(x.size(0), -1)
    x = self.dropout(x)
    x = self.fc1(x)
    return x

我一直在使用形状为 128*128 的图像运行训练循环，并且层数较少，因此我能够计算全连接层的输入（展平）大小，但是如果我有多个无法手动计算的层怎么办。如何使我的神经网络的全连接层输入兼容所有形状的图像？

Answer 1

您可以使用

hook

来获取

nn.Linear

之前一层的输出。请参考下面的源代码：

 class CNN(nn.Module):
    def __init__(self, num_classes):
        super(CNN, self).__init__()
        self.conv1 = nn.Conv2d(
            in_channels=3, out_channels=16, kernel_size=3, stride=1, padding=0)
        self.conv2 = nn.Conv2d(
            in_channels=16, out_channels=32, kernel_size=3, stride=1, padding=0)
        self.conv2.register_forward_hook(self._hook_conv2)
        self.fc1 = None
        self.fc2 = nn.Linear(128, num_classes)

    def _hook_conv2(self, module, input, output):
        print(output.size())
        self.fc1 = nn.Linear(output.size(1) * output.size(2) * output.size(3), 128)

    def forward(self, x):
        x = self.conv1(x)
        x = self.conv2(x)
        x = x.view(x.size(0), -1)
        if self.fc1 is not None:
            x = self.fc1(x)
        x = self.fc2(x)
        return x