lenet5.py

import torch.nn as nn
import torch.nn.functional as func

class lenet(nn.Module):
    '''
    input: 3x32x32 image
    output: 10 class probability
    '''
    def __init__(self):
        super(lenet, self).__init__()
        self.conv1 = nn.Conv2d(3,6,5) #c1:featuremaps 6@28x28 #output = (input-filter)/stride + 1, #filter:5size
        self.conv2 = nn.Conv2d(6,16,5) #c3:feature_maps 16@10x10
        self.maxPool = nn.MaxPool2d(2,2) #subsampling 1/2size
        self.fc1 = nn.Linear(16*5*5,120) #f5:layer120
        self.fc2 = nn.Linear(120,84) #f6:layer84
        self.fc3 = nn.Linear(84,10) #output:10 class

    def forward(self,x):
        x = self.maxPool(func.relu(self.conv1(x)))
        x = self.maxPool(func.relu(self.conv2(x)))
        x = x.view(-1, 16*5*5) #flattens #tensor 형태 변환, 일렬로 16*5*5형태로 만듦
        x = func.relu(self.fc1(x))
        x = func.relu(self.fc2(x))
        x = self.fc3(x)
        return x