【无标题】

​​​​​​​

AECRNET.py

import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.nn import init
import functools
from deconv import FastDeconv
from DCNv2.dcn_v2 import DCN

# 默认卷积，不改变维度大小，仅仅是为了提取特征
def default_conv(in_channels, out_channels, kernel_size, bias=True):
    return nn.Conv2d(in_channels, out_channels, kernel_size, padding=(kernel_size // 2), bias=bias)


class PALayer(nn.Module):
    def __init__(self, channel):
        super(PALayer, self).__init__()
        self.pa = nn.Sequential(
            nn.Conv2d(channel, channel // 8, 1, padding=0, bias=True),
            nn.ReLU(inplace=True),
            nn.Conv2d(channel // 8, 1, 1, padding=0, bias=True),
            nn.Sigmoid()
        )

    def forward(self, x):
        y = self.pa(x)
        return x * y


class CALayer(nn.Module):
    def __init__(self, channel):
        super(CALayer, self).__init__()
        self.avg_pool = nn.AdaptiveAvgPool2d(1)
        self.ca = nn.Sequential(
            nn.Conv2d(channel, channel // 8, 1, padding=0, bias=True),
            nn.ReLU(inplace=True),
            nn.Conv2d(channel // 8, channel, 1, padding=0, bias=True),
            nn.Sigmoid()
        )

    def forward(self, x):
        y = self.avg_pool(x)
        y = self.ca(y)
        return x * y


class DehazeBlock(nn.Module):
    def __init__(self, conv, dim, kernel_size, ):
        super(DehazeBlock, self).__init__()
        self.conv1 = conv(dim, dim, kernel_size, bias=True)
        self.act1 = nn.ReLU(inplace=True)
        self.conv2 = conv(dim, dim, kernel_size, bias=True)
        self.calayer = CALayer(dim)
        self.palayer = PALayer(dim)

    def forward(self, x):
        res = self.act1(self.conv1(x))
        res = res + x
        res = self.conv2(res)
        res = self.calayer(res)
        res = self.palayer(res)
        res += x
        return res


class DCNBlock(nn.Module):
    def __init__(self, in_channel, out_channel):
        super(DCNBlock, self).__init__()
        self.dcn = DCN(in_channel, out_channel, kernel_size=(3,3), stride=1, padding=1).cuda()
    def forward(self, x):
        return self.dcn(x)

# 自适应混合
class Mix(nn.Module):
    def __init__(self, m=-0.80):
        super(Mix, self).__init__()
        w = torch.nn.Parameter(torch.FloatTensor([m]), requires_grad=True)
        w = torch.nn.Parameter(w, requires_grad=True)
        self.w = w
        self.mix_block = nn.Sigmoid()

    def forward(self, fea1, fea2):
        mix_factor = self.mix_block(self.w)
        out = fea1 * mix_factor.expand_as(fea1) + fea2 * (1 - mix_factor.expand_as(fea2))
        return out

class Dehaze(nn.Module):
    def __init__(self, input_nc, output_nc, ngf=64, use_dropout=False, padding_type='reflect'):
        super(Dehaze, self).__init__()

        ###### downsample
        self.down1 = nn.Sequential(nn.ReflectionPad2d(3),
                                   nn.Conv2d(input_nc, ngf, kernel_size=7, padding=0),
                                   nn.ReLU(True))
        self.down2 = nn.Sequential(nn.Conv2d(ngf, ngf*2, kernel_size=3, stride=2, padding=1),
                                   nn.ReLU(True))
        self.down3 = nn.Sequential(nn.Conv2d(ngf*2, ngf*4, kernel_size=3, stride=2, padding=1),
                                   nn.ReLU(True))

        ###### FFA blocks
        self.block = DehazeBlock(default_conv, ngf * 4, 3)

        ###### upsample
        self.up1 = nn.Sequential(nn.ConvTranspose2d(ngf*4, ngf*2, kernel_size=3, stride=2, padding=1, output_padding=1),
                                 nn.ReLU(True))
        self.up2 = nn.Sequential(nn.ConvTranspose2d(ngf*2, ngf, kernel_size=3, stride=2, padding=1, output_padding=1),
                                 nn.ReLU(True))
        self.up3 = nn.Sequential(nn.ReflectionPad2d(3),
                                 nn.Conv2d(ngf, output_nc, kernel_size=7, padding=0),
                                 nn.Tanh())


        self.dcn_block = DCNBlock(256, 256)

        self.deconv = FastDeconv(3, 3, kernel_size=3, stride=1, padding=1)

        self.mix1 = Mix(m=-1)
        self.mix2 = Mix(m=-0.6)

    def forward(self, input):

        x_deconv = self.deconv(input) # preprocess

        x_down1 = self.down1(x_deconv) # [bs, 64, 256, 256]
        x_down2 = self.down2(x_down1) # [bs, 128, 128, 128]
        x_down3 = self.down3(x_down2) # [bs, 256, 64, 64]

        x1 = self.block(x_down3)
        x2 = self.block(x1)
        x3 = self.block(x2)
        x4 = self.block(x3)
        x5 = self.block(x4)
        x6 = self.block(x5)

        x_dcn1 = self.dcn_block(x6)
        x_dcn2 = self.dcn_block(x_dcn1)

        x_out_mix = self.mix1(x_down3, x_dcn2)
        x_up1 = self.up1(x_out_mix) # [bs, 128, 128, 128]
        x_up1_mix = self.mix2(x_down2, x_up1)
        x_up2 = self.up2(x_up1_mix) # [bs, 64, 256, 256] 
        out = self.up3(x_up2) # [bs,  3, 256, 256]

        return out

CR.py 

import torch.nn as nn
import torch
from torch.nn import functional as F
import torch.nn.functional as fnn
from torch.autograd import Variable
import numpy as np
from torchvision import models

class Vgg19(torch.nn.Module):
    def __init__(self, requires_grad=False):
        super(Vgg19, self).__init__()
        vgg_pretrained_features = models.vgg19(pretrained=True).features
        self.slice1 = torch.nn.Sequential()
        self.slice2 = torch.nn.Sequential()
        self.slice3 = torch.nn.Sequential()
        self.slice4 = torch.nn.Sequential()
        self.slice5 = torch.nn.Sequential()
        for x in range(2):
            self.slice1.add_module(str(x), vgg_pretrained_features[x])
        for x in range(2, 7):
            self.slice2.add_module(str(x), vgg_pretrained_features[x])
        for x in range(7, 12):
            self.slice3.add_module(str(x), vgg_pretrained_features[x])
        for x in range(12, 21):
            self.slice4.add_module(str(x), vgg_pretrained_features[x])
        for x in range(21, 30):
            self.slice5.add_module(str(x), vgg_pretrained_features[x])
        if not requires_grad:
            for param in self.parameters():
                param.requires_grad = False

    def forward(self, X):
        h_relu1 = self.slice1(X)
        h_relu2 = self.slice2(h_relu1) 
        h_relu3 = self.slice3(h_relu2)
        h_relu4 = self.slice4(h_relu3)
        h_relu5 = self.slice5(h_relu4) 
        return [h_relu1, h_relu2, h_relu3, h_relu4, h_relu5]

class ContrastLoss(nn.Module):
    def __init__(self, ablation=False):

        super(ContrastLoss, self).__init__()
        self.vgg = Vgg19().cuda()
        self.l1 = nn.L1Loss()
        self.weights = [1.0/32, 1.0/16, 1.0/8, 1.0/4, 1.0]
        self.ab = ablation

    def forward(self, a, p, n):
        a_vgg, p_vgg, n_vgg = self.vgg(a), self.vgg(p), self.vgg(n)
        loss = 0

        d_ap, d_an = 0, 0
        for i in range(len(a_vgg)):
            d_ap = self.l1(a_vgg[i], p_vgg[i].detach())
            if not self.ab:
                d_an = self.l1(a_vgg[i], n_vgg[i].detach())
                contrastive = d_ap / (d_an + 1e-7)
            else:
                contrastive = d_ap

            loss += self.weights[i] * contrastive
        return loss