如何从原始图像尺寸计算内核尺寸?

How to calculate kernel dimensions from original image dimensions?

https://github.com/kuangliu/pytorch-cifar/blob/master/models/resnet.py

根据阅读 https://www.cs.toronto.edu/~kriz/cifar.html,cifar 数据集由每张 32x32 维度的图像组成。

我对代码的理解:

self.conv1 = nn.Conv2d(3, 6, 5)
self.conv2 = nn.Conv2d(6, 16, 5)
self.fc1   = nn.Linear(16*5*5, 120)

是:

self.conv1 = nn.Conv2d(3, 6, 5) # 3 channels in, 6 channels out ,  kernel size of 5
self.conv2 = nn.Conv2d(6, 16, 5) # 6 channels in, 16 channels out ,  kernel size of 5
self.fc1   = nn.Linear(16*5*5, 120) # 16*5*5 in features , 120 ouot feature

来自 resnet.py 以下 :

self.fc1   = nn.Linear(16*5*5, 120)

来自 http://cs231n.github.io/convolutional-networks/ 的内容如下:

Summary. To summarize, the Conv Layer:

Accepts a volume of size W1×H1×D1 Requires four hyperparameters: Number of filters K, their spatial extent F, the stride S, the amount of zero padding P. Produces a volume of size W2×H2×D2 where: W2=(W1−F+2P)/S+1 H2=(H1−F+2P)/S+1 (i.e. width and height are computed equally by symmetry) D2=K With parameter sharing, it introduces F⋅F⋅D1 weights per filter, for a total of (F⋅F⋅D1)⋅K weights and K biases. In the output volume, the d-th depth slice (of size W2×H2) is the result of performing a valid convolution of the d-th filter over the input volume with a stride of S, and then offset by d-th bias.

据此我试图了解训练图像尺寸 32x32(1024 像素)如何转换为特征图(16*5*5 -> 400)作为 nn.Linear(16*5*5, 120)

https://pytorch.org/docs/stable/nn.html#torch.nn.Conv2d可以看出默认stride为1,padding为0。

从 32*32 的图像尺寸得到 16*5*5 的步骤是什么?可以从上述步骤导出 16*5*5 吗?

从以上步骤如何计算spatial extent ?

更新:

源代码:

'''LeNet in PyTorch.'''
import torch.nn as nn
import torch.nn.functional as F

class LeNet(nn.Module):
    def __init__(self):
        super(LeNet, self).__init__()
        self.conv1 = nn.Conv2d(3, 6, 5)
        self.conv2 = nn.Conv2d(6, 16, 5)
        self.fc1   = nn.Linear(16*5*5, 120)
        self.fc2   = nn.Linear(120, 84)
        self.fc3   = nn.Linear(84, 10)

    def forward(self, x):
        out = F.relu(self.conv1(x))
        out = F.max_pool2d(out, 2)
        out = F.relu(self.conv2(out))
        out = F.max_pool2d(out, 2)
        out = out.view(out.size(0), -1)
        out = F.relu(self.fc1(out))
        out = F.relu(self.fc2(out))
        out = self.fc3(out)
        return out

取自https://github.com/kuangliu/pytorch-cifar/blob/master/models/lenet.py

我的理解是卷积运算应用于每个内核的图像数据。因此,如果设置了 5 个内核,则对生成 5 维图像表示的数据应用 5 个卷积。

您在问题中没有提供足够的信息(参见 )。

但是,如果我不得不猜测,那么您的卷积层之间有两个池化层(步幅为 2):

  • 输入大小 32x32(3 个通道)
  • conv1 输出大小 28x28(6 通道):conv 没有填充且内核大小为 5,将输入大小减少 4。
  • 步幅为 2 的池化层,输出大小为 14x14(6 个通道)。
  • conv2 输出大小 10x10(16 个通道)
  • 另一个步幅为 2 的池化层,输出大小为 5x5(16 个通道)
  • 一个全连接层 (nn.Linear) 将所有 5x5x16 输入连接到所有 120 个输出。

可以找到更全面的感受野估计指南here