CV中的注意力机制(一)——SENet(通道注意力)

SENet

实现步骤

1. 对输入进来的特征层进行全局平均池化。
2. 然后进行两次全连接，第一次全连接神经元个数较少，第二次全连接神经元个数和输入特征层相同。
3. 在完成两次全连接后，再取一次Sigmoid将值固定到0-1之间，此时获得了输入特征层每一个通道的权值（0-1之间）。
4. 将获得的权值乘上原输入特征层即可。

流程图示

结合ResNet网络实战

左图为原始的residual模块，右图为加入SE的residual模块。

测试

uhhh，自己的本子是3070的，就不在ImageNet1k上做测试了，还是在ImageNet-mini浅试一下吧。先放出来最终的效果：
从图中很明显的可以看出来，加入了SE模块后的ResNet大大降低了错误率，加入SE模块后，计算量会增大很多，这也是为什么虽然SE模块能提高准确率，但是没有被广泛应用的最大的原因。废话不多说，上代码，注释就免了。。。。。我还是相信各位大佬的实力的。

SE模块

import torch
import torch.nn as nn


class se_block(nn.Module):
    def __init__(self, channels, ratio=16):
        super(se_block, self).__init__()
        self.avg_pool = nn.AdaptiveAvgPool2d((1, 1))
        self.fc = nn.Sequential(
            nn.Linear(channels, channels // ratio, bias=False),
            nn.ReLU(inplace=True),
            nn.Linear(channels // ratio, channels, bias=False),
            nn.Sigmoid()
        )

    def forward(self, x):
        b, c, _, _ = x.size()
        y = self.avg_pool(x).view(b, c)
        y = self.fc(y).view(b, c, 1, 1)
        return x * y.expand_as(x)

模型代码

import torch
import torch.nn as nn
from module.se_module import se_block


class SEBottleneck(nn.Module):
    expansion = 4
    def __init__(self, in_planes, out_planes, stride=(1, 1),
                 downsample=None, reduction=16):
        super(SEBottleneck, self).__init__()
        self.conv1 = nn.Conv2d(in_planes, out_planes, kernel_size=(1, 1), bias=False)
        self.bn1 = nn.BatchNorm2d(out_planes)
        self.conv2 = nn.Conv2d(out_planes, out_planes, kernel_size=(3, 3), stride=stride,
                               padding=1, bias=False)
        self.bn2 = nn.BatchNorm2d(out_planes)
        self.conv3 = nn.Conv2d(out_planes, out_planes * 4, kernel_size=(1, 1), bias=False)
        self.bn3 = nn.BatchNorm2d(out_planes * 4)
        self.relu = nn.ReLU(inplace=True)
        self.se = se_block(out_planes * 4, ratio=reduction)
        self.downsample = downsample
        self.stride = stride

    def forward(self, x):
        residual = x

        out = self.conv1(x)
        out = self.bn1(out)
        out = self.relu(out)

        out = self.conv2(out)
        out = self.bn2(out)
        out = self.relu(out)

        out = self.conv3(out)
        out = self.bn3(out)
        out = self.se(out)

        if self.downsample is not None:
            residual = self.downsample(x)
        out += residual
        out = self.relu(out)
        return out


class ResNet(nn.Module):

    def __init__(self, block, blocks_num, num_classes=1000, include_top=True):
        super(ResNet, self).__init__()
        self.include_top = include_top
        self.in_channel = 64
        self.conv1 = nn.Conv2d(3, self.in_channel, kernel_size=(7, 7), stride=(2, 2),
                               padding=3, bias=False)
        self.bn1 = nn.BatchNorm2d(self.in_channel)
        self.relu = nn.ReLU(inplace=True)
        self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
        self.layer1 = self._make_layer(block, 64, blocks_num[0])
        self.layer2 = self._make_layer(block, 128, blocks_num[1], stride=2)
        self.layer3 = self._make_layer(block, 256, blocks_num[2], stride=2)
        self.layer4 = self._make_layer(block, 512, blocks_num[3], stride=2)
        if self.include_top:
            self.avgpool = nn.AdaptiveAvgPool2d((1, 1))  # output size = (1, 1)
            self.fc = nn.Linear(512 * block.expansion, num_classes)

        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
            elif isinstance(m, nn.BatchNorm2d):
                nn.init.constant_(m.weight, 1)
                nn.init.constant_(m.bias, 0)

    def _make_layer(self, block, channel, block_num, stride=(1, 1)):
        downsample = None
        if stride != 1 or self.in_channel != channel * block.expansion:
            downsample = nn.Sequential(
                nn.Conv2d(self.in_channel, channel * block.expansion, kernel_size=(1, 1),
                          stride=stride, bias=False),
                nn.BatchNorm2d(channel * block.expansion))

        layers = []
        layers.append(block(self.in_channel,
                            channel,
                            downsample=downsample,
                            stride=stride))
        self.in_channel = channel * block.expansion

        for _ in range(1, block_num):
            layers.append(block(self.in_channel,
                                channel))

        return nn.Sequential(*layers)

    def forward(self, x):
        x = self.conv1(x)
        x = self.bn1(x)
        x = self.relu(x)
        x = self.maxpool(x)

        x = self.layer1(x)
        x = self.layer2(x)
        x = self.layer3(x)
        x = self.layer4(x)

        if self.include_top:
            x = self.avgpool(x)
            x = torch.flatten(x, 1)
            x = self.fc(x)

        return x

def se_resnet50(num_classes=1000, include_top=True):
    return ResNet(SEBottleneck, [3, 4, 6, 3], num_classes=num_classes, include_top=include_top)

训练及测试

import os
import torch
import numpy as np
import torch.nn as nn
from tqdm import tqdm
import torch.optim as optim
import matplotlib.pylab as plt
from module.se_resnet import se_resnet50
from module.resnet import resnet50
from data import _main

def main():
    device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
    print('Using {} device for train.'.format(device))
    train_num, train_loader, val_num, val_loader = _main()
    weights_path = r'seresnet50-60a8950a85b2b.pkl'
    model = se_resnet50(num_classes=100)
    #
    # weights_path = r'resnet50-0676ba61.pth'
    # model = resnet50(num_classes=100)

    pre_weights = torch.load(weights_path, map_location=device)
    pre_dict = {k: v for k, v in pre_weights.items() if model.state_dict()[k].numel() == v.numel()}
    missing_keys, unexpected_keys = model.load_state_dict(pre_dict, strict=False)
    model.to(device)
    loss_function = nn.CrossEntropyLoss()
    params =[p for p in model.parameters() if p.requires_grad]
    # optimizer = optim.SGD(params, lr=0.3, momentum=0.9, weight_decay=1e-4)
    optimizer = optim.Adam(params, lr=2e-4)
    epochs = 40
    save_path = 'se_resnet50.pkl'
    # save_path = 'resnet50.pth'
    best_acc = 0.0
    train_steps = len(train_loader)

    for epoch in range(epochs):
        model.train()
        running_loss = 0.0
        train_bar = tqdm(train_loader)
        for step, data in enumerate(train_bar):
            images, labels = data
            optimizer.zero_grad()
            outputs = model(images.to(device))
            loss = loss_function(outputs, labels.to(device))
            loss.backward()
            optimizer.step()

            running_loss += loss.item()

            train_bar.desc = 'train epoch[{}/{}] loss:{:.3f}'.format(epoch + 1, epochs, loss)
        loss_list.append(running_loss / train_steps)
        model.eval()
        acc = 0.0
        with torch.no_grad():
            val_bar = tqdm(val_loader)
            for val_data in val_bar:
                val_images, val_labels = val_data
                outputs = model(val_images.to(device))
                _, predice_y = torch.max(outputs, dim=1)
                acc += torch.eq(predice_y, val_labels.to(device)).sum().item()
            val_acc = acc / val_num
            print('[epoch %d] training loss:%.3f val_accuracy:%.3f'
                  % (epoch + 1, running_loss / train_steps, val_acc))
            if val_acc > best_acc:
                best_acc = val_acc
                torch.save(model.state_dict(), save_path)

    print('Finished Training!')
    print('Best accuracy is {}'.format(best_acc))
    
if __name__ == '__main__':
    main()

欢迎转载与提问！！！！！！！！