CNN学习系列：骨干网络学习之AlexNet

论文地址：ImageNet Classification with Deep Convolutional Neural Networks

网络结构为

使用pytorch复现该网络

mport torch.nn as nn
from torch.nn import functional as F
from torchsummary import summary


# pytorch默认的实现方式为torchvision.models.alexnet()
class AlexNet(nn.Module):
    def __init__(self, num_classes=1000):
        super(AlexNet, self).__init__()
        self.conv1 = nn.Conv2d(in_channels=3, out_channels=96, kernel_size=(11, 11), stride=(4, 4), padding=(1, 1))
        self.conv2 = nn.Conv2d(in_channels=96, out_channels=256, kernel_size=(5, 5), groups=2, padding=(2, 2))
        self.conv3 = nn.Conv2d(in_channels=256, out_channels=384, padding=(1, 1), kernel_size=(3, 3))
        self.conv4 = nn.Conv2d(in_channels=384, out_channels=384, padding=(1, 1), kernel_size=(3, 3))
        self.conv5 = nn.Conv2d(in_channels=384, out_channels=256, padding=(1, 1), kernel_size=(3, 3))

        self.fc1 = nn.Linear(in_features=6 * 6 * 256, out_features=4096)
        self.fc2 = nn.Linear(in_features=4096, out_features=4096)
        self.fc3 = nn.Linear(in_features=4096, out_features=num_classes)

    def forward(self, t):
        t = F.relu(self.conv1(t))
        t = F.max_pool2d(t, kernel_size=(3, 3), stride=(2, 2))
        t = F.relu(self.conv2(t))
        t = F.max_pool2d(t, kernel_size=(3, 3), stride=(2, 2))
        t = F.relu(self.conv3(t))
        t = F.relu(self.conv4(t))
        t = F.relu(self.conv5(t))
        t = F.max_pool2d(t, kernel_size=(3, 3), stride=(2, 2))

        t = t.view(-1, 6 * 6 * 256)
        t = F.relu(self.fc1(t))
        t = F.relu(self.fc2(t))
        t = F.relu(self.fc3(t))

        return t


if __name__ == '__main__':
    alex_net = AlexNet()
    alex_net = alex_net.cuda()
    summary(alex_net, (3, 227, 227))