net_class.py

# -*- encoding: utf-8 -*-
import torch.nn as nn
import torch


class SiameseNetwork(nn.Module):
    def __init__(self):
        super(SiameseNetwork, self).__init__()  # TODO解决初始化的问题
        # self.cnn1 = nn.Sequential(
        #     nn.Conv2d(1, 20, kernel_size=5),
        #     nn.ReLU(inplace=True),
        #     # nn.MaxPool2d(2, stride=2),
        #     nn.Conv2d(20, 50, kernel_size=5),
        #     # nn.MaxPool2d(2, stride=2),
        #     nn.ReLU(inplace=True))
        self.cnn1 = nn.Conv2d(2, 10, kernel_size=5)
        self.pool1 = nn.MaxPool2d(2,stride=2)
        self.cnn2 = nn.Conv2d(10, 50, kernel_size=3)
        self.pool2 = nn.MaxPool2d(2,stride=2)
        self.cnn3 = nn.Conv2d(50, 250, kernel_size=3)
        self.pool3 = nn.MaxPool2d(2,stride=2)
        self.cnn4 = nn.Conv2d(250, 750, kernel_size=3)
        self.pool4 = nn.MaxPool2d(2,stride=2)
        self.cnn5 = nn.Conv2d(750, 1250, kernel_size=3)

        self.fc1 = nn.Linear(5000, 3000)
        self.fc2 = nn.Linear(3000, 1000)
        self.fc3 = nn.Linear(1000, 100)
        self.fc4 = nn.Linear(100, 10)
        self.bn1 = nn.BatchNorm2d(10)
        self.bn2 = nn.BatchNorm2d(50)
        self.bn3 = nn.BatchNorm2d(250)
        self.bn4 = nn.BatchNorm2d(750)
        self.bn5 = nn.BatchNorm2d(1250)
        self.relu = nn.ReLU(inplace=True)
        '''
        self.fc1 = nn.Linear(1800, 200)#30*30*5#1800
        self.fc2 = nn.Linear(200, 15)
        '''

    def forward(self, x):
        out = self.bn1(self.relu(self.cnn1(x)))
        out = self.pool1(out)
        out = self.bn2(self.relu(self.cnn2(out)))
        # out = self.pool2(out)
        out = self.bn3(self.relu(self.cnn3(out)))
        # out = self.pool3(out)
        out = self.bn4(self.relu(self.cnn4(out)))
        out = self.pool3(out)
        out = self.bn5(self.relu(self.cnn5(out)))
        # out = self.pool3(out)
        # out = self.bn4(self.cnn4(out))
        # out = self.pool4(out)
        #print out.size()
        # relu1_out = F.max_pool2d(relu1_out, 2)
        out = out.view(-1, 5000)  # 把矩阵拉成向量
        out = self.relu(self.fc1(out))
        out = self.relu(self.fc2(out))
        out = self.relu(self.fc3(out))
        out = self.fc4(out)#relu remove


        '''
        # relu1_out = F.max_pool2d(relu1_out, 2)
        out = out.view(-1,4500)  # 把矩阵拉成向量
        out = F.relu(self.fc0(out))
        out = F.relu(self.fc01(out))
        out = F.relu(self.fc02(out))
        out = F.relu(self.fc03(out))
        out = F.relu(self.fc04(out))
        out = F.relu(self.fc05(out))
        #out = self.fc2(out)
        '''
        return out
'''
    def forward(self, input1, input2):
        output1 = self.forward_once(input1)
        output2 = self.forward_once(input2)
        # print(output1)
        return output1, output2
'''