resnet.py

import torch.nn as nn
import torch.utils.model_zoo as model_zoo
import torch.nn.functional as F

__all__ = ['ResNet', 'resnet18', 'resnet34', 'resnet50', 'resnet101',
           'resnet152']


model_urls = {
    'resnet18': 'https://download.pytorch.org/models/resnet18-5c106cde.pth',
    'resnet34': './pretrain/resnet34-333f7ec4.pth',
    'resnet50': './pretrain/resnet50-19c8e357.pth',
    'resnet101': './pretrain/resnet101-5d3b4d8f.pth',
    'resnet152': './pretrain/resnet152-b121ed2d.pth',
}


def conv3x3(in_planes, out_planes, stride=1):
    """3x3 convolution with padding"""
    return nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride,
                     padding=1, bias=False)


def conv1x1(in_planes, out_planes, stride=1):
    """1x1 convolution"""
    return nn.Conv2d(in_planes, out_planes, kernel_size=1, stride=stride, bias=False)



class SepConv(nn.Module):

    def __init__(self, channel_in, channel_out, kernel_size=3, stride=2, padding=1, affine=True):
        #   depthwise and pointwise convolution, downsample by 2
        super(SepConv, self).__init__()
        self.op = nn.Sequential(
            nn.Conv2d(channel_in, channel_in, kernel_size=kernel_size, stride=stride, padding=padding, groups=channel_in, bias=False),
            nn.Conv2d(channel_in, channel_in, kernel_size=1, padding=0, bias=False),
            nn.BatchNorm2d(channel_in, affine=affine),
            nn.ReLU(inplace=False),
            nn.Conv2d(channel_in, channel_in, kernel_size=kernel_size, stride=1, padding=padding, groups=channel_in, bias=False),
            nn.Conv2d(channel_in, channel_out, kernel_size=1, padding=0, bias=False),
            nn.BatchNorm2d(channel_out, affine=affine),
            nn.ReLU(inplace=False),
        )

    def forward(self, x):
        return self.op(x)


class BasicBlock(nn.Module):
    expansion = 1

    def __init__(self, inplanes, planes, stride=1, downsample=None):
        super(BasicBlock, self).__init__()
        self.conv1 = conv3x3(inplanes, planes, stride)
        self.bn1 = nn.BatchNorm2d(planes)
        self.relu = nn.ReLU(inplace=True)
        self.conv2 = conv3x3(planes, planes)
        self.bn2 = nn.BatchNorm2d(planes)
        self.downsample = downsample
        self.stride = stride

    def forward(self, x):
        identity = x

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

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

        if self.downsample is not None:
            identity = self.downsample(x)

        out += identity
        out = self.relu(out)

        return out


class Bottleneck(nn.Module):
    expansion = 4

    def __init__(self, inplanes, planes, stride=1, downsample=None):
        super(Bottleneck, self).__init__()
        self.conv1 = conv1x1(inplanes, planes)
        self.bn1 = nn.BatchNorm2d(planes)
        self.conv2 = conv3x3(planes, planes, stride)
        self.bn2 = nn.BatchNorm2d(planes)
        self.conv3 = conv1x1(planes, planes * self.expansion)
        self.bn3 = nn.BatchNorm2d(planes * self.expansion)
        self.relu = nn.ReLU(inplace=True)
        self.downsample = downsample
        self.stride = stride

    def forward(self, x):
        identity = 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)

        if self.downsample is not None:
            identity = self.downsample(x)

        out += identity
        out = self.relu(out)

        return out

class ResNet(nn.Module):
    def __init__(self, block, layers, num_classes=100, zero_init_residual=False, align="CONV"):
        super(ResNet, self).__init__()
        self.inplanes = 64
        self.align = align
        #   reduce the kernel-size and stride of ResNet on cifar datasets.
        self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=1, padding=3,
                               bias=False)
        self.bn1 = nn.BatchNorm2d(64)
        self.relu = nn.ReLU(inplace=True)
        #   remove maxpooling layer for ResNet on cifar datasets.
        #   self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
        self.layer1 = self._make_layer(block, 64, layers[0])
        self.layer2 = self._make_layer(block, 128, layers[1], stride=2)
        self.layer3 = self._make_layer(block, 256, layers[2], stride=2)
        self.layer4 = self._make_layer(block, 512, layers[3], stride=2)

        self.auxiliary1 = nn.Sequential(
            SepConv(
                channel_in=64 * block.expansion,
                channel_out=128 * block.expansion
            ),
            SepConv(
                channel_in=128 * block.expansion,
                channel_out=256 * block.expansion
            ),
            SepConv(
                channel_in=256 * block.expansion,
                channel_out=512 * block.expansion
            ),
            nn.AvgPool2d(4, 4)
        )

        self.auxiliary2 = nn.Sequential(
            SepConv(
                channel_in=128 * block.expansion,
                channel_out=256 * block.expansion,
            ),
            SepConv(
                channel_in=256 * block.expansion,
                channel_out=512 * block.expansion,
            ),
            nn.AvgPool2d(4, 4)
        )
        self.auxiliary3 = nn.Sequential(
            SepConv(
                channel_in=256 * block.expansion,
                channel_out=512 * block.expansion,
            ),
            nn.AvgPool2d(4, 4)
        )
        self.auxiliary4 = nn.AvgPool2d(4, 4)

        self.fc_o = nn.Linear(512 , num_classes)#25088 #2048
        #print(512 * block.expansion)
        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)

        # Zero-initialize the last BN in each residual branch,
        # so that the residual branch starts with zeros, and each residual block behaves like an identity.
        # This improves the model by 0.2~0.3% according to https://arxiv.org/abs/1706.02677
        if zero_init_residual:
            for m in self.modules():
                if isinstance(m, Bottleneck):
                    nn.init.constant_(m.bn3.weight, 0)
                elif isinstance(m, BasicBlock):
                    nn.init.constant_(m.bn2.weight, 0)

    def _make_layer(self, block, planes, blocks, stride=1):
        downsample = None
        if stride != 1 or self.inplanes != planes * block.expansion:
            downsample = nn.Sequential(
                conv1x1(self.inplanes, planes * block.expansion, stride),
                nn.BatchNorm2d(planes * block.expansion),
            )

        layers = []
        layers.append(block(self.inplanes, planes, stride, downsample))
        self.inplanes = planes * block.expansion
        for _ in range(1, blocks):
            layers.append(block(self.inplanes, planes))

        return nn.Sequential(*layers)

    def forward(self, x):
        feature_list = []
        x = self.conv1(x)
        x = self.bn1(x)
        x = self.relu(x)
        x = self.layer1(x)
        feature_list.append(x)
        x = self.layer2(x)
        feature_list.append(x)
        x = self.layer3(x)
        feature_list.append(x)
        x = self.layer4(x)
        feature_list.append(x)

        out1_feature = self.auxiliary1(feature_list[0]).view(x.size(0), -1)
        out2_feature = self.auxiliary2(feature_list[1]).view(x.size(0), -1)
        out3_feature = self.auxiliary3(feature_list[2]).view(x.size(0), -1)
        out4_feature = self.auxiliary4(feature_list[3]).view(x.size(0), -1)
        #print(out4_feature.shape)
        out = self.fc_o(out4_feature)
        #print(out.shape)
        feat_list = [out4_feature, out3_feature, out2_feature, out1_feature]
        for index in range(len(feat_list)):
            feat_list[index] = F.normalize(feat_list[index], dim=1)
        if self.training:
            return out, feat_list
        else:
            return out
def resnet18(pretrained=False, **kwargs):
    """Constructs a ResNet-18 model.

    Args:
        pretrained (bool): If True, returns a model pre-trained on ImageNet
    """
    model = ResNet(BasicBlock, [2, 2, 2, 2], **kwargs)
    if pretrained:
        model_dict = model.state_dict()
        pretrained_dict = model_zoo.load_url(model_urls['resnet18'])
        pretrained_dict = {k: v for k, v in pretrained_dict.items() if k in model_dict}
        #print(pretrained_dict)
        model_dict.update(pretrained_dict)
        model.load_state_dict(model_dict)
    return model


def resnet34(pretrained=False, **kwargs):
    """Constructs a ResNet-34 model.

    Args:
        pretrained (bool): If True, returns a model pre-trained on ImageNet
    """
    model = ResNet(BasicBlock, [3, 4, 6, 3], **kwargs)
    if pretrained:
        model.load_state_dict(model_zoo.load_url(model_urls['resnet34']))
    return model


def resnet50(pretrained=False, **kwargs):
    """Constructs a ResNet-50 model.

    Args:
        pretrained (bool): If True, returns a model pre-trained on ImageNet
    """
    model = ResNet(Bottleneck, [3, 4, 6, 3], **kwargs)
    if pretrained:
        model_dict = model.state_dict()
        pretrained_dict = model_zoo.load_url(model_urls['resnet50'])
        pretrained_dict = {k: v for k, v in pretrained_dict.items() if k in model_dict}
        #print(pretrained_dict)
        model_dict.update(pretrained_dict)
        model.load_state_dict(model_dict)
        # model.backbone.conv1 = nn.Conv2d(in_channel, 64, kernel_size=7, stride=2, padding=3, bias=False)
        # model.classifier = ResNet(2048, out_channel)
        # model.load_state_dict(model_zoo.load_url(model_urls['resnet50']))
    return model


def resnet101(pretrained=False, **kwargs):
    """Constructs a ResNet-101 model.

    Args:
        pretrained (bool): If True, returns a model pre-trained on ImageNet
    """
    model = ResNet(Bottleneck, [3, 4, 23, 3], **kwargs)
    if pretrained:
        model.load_state_dict(model_zoo.load_url(model_urls['resnet101']))
    return model


def resnet152(pretrained=False, **kwargs):
    """Constructs a ResNet-152 model.

    Args:
        pretrained (bool): If True, returns a model pre-trained on ImageNet
    """
    model = ResNet(Bottleneck, [3, 8, 36, 3], **kwargs)
    if pretrained:
        model.load_state_dict(model_zoo.load_url(model_urls['resnet152']))

    return model

import torch.optim as optim
import os, torch
from loss import SupConLoss

if __name__ == '__main__':
    h = torch.rand((8, 3, 256, 256)).cuda()
    l = torch.tensor([0,3,2,3]).cuda()
    #contra_criterion = SupConLoss()
    #clean_criterion = nn.CrossEntropyLoss(reduce=True)
    model =resnet18(pretrained=False).cuda()
    #optimizer = optim.SGD(model.parameters(), lr=1e-3, momentum=0.9, weight_decay=1e-4)
    out, features = model(h)
    f1, f2 = torch.split(features[2], [4, 4], dim=0)
    features = torch.cat([f1.unsqueeze(1), f2.unsqueeze(1)], dim=1)
    #loss = contra_criterion(features, labels=l) * 1e-1
    print('================')
    #torch.save(model.state_dict(), os.path.join('./epoch_{}.pt'.format(1)))