Spaces:
Build error
Build error
| """ | |
| This file contains a code from ***MakeItTalk*** | |
| This code is originally from the project name MakeItTalk from "" at MakeItTalk repository | |
| Link :https://github.com/yzhou359/MakeItTalk | |
| """ | |
| import torch | |
| import torch.nn as nn | |
| #import torch.nn.parallel | |
| #from torch.autograd import Variable | |
| import torch.nn.functional as F | |
| #from torchvision import models | |
| #import torch.utils.model_zoo as model_zoo | |
| #from torch.nn import init | |
| import os | |
| import numpy as np | |
| class ResUnetGenerator(nn.Module): | |
| """ | |
| Main Image2Image Translation Network | |
| """ | |
| def __init__(self, input_nc, output_nc, num_downs, ngf=64, | |
| norm_layer=nn.BatchNorm2d, use_dropout=False): | |
| super(ResUnetGenerator, self).__init__() | |
| # construct unet structure | |
| unet_block = ResUnetSkipConnectionBlock(ngf * 8, ngf * 8, input_nc=None, submodule=None, norm_layer=norm_layer, | |
| innermost=True) | |
| for i in range(num_downs - 5): | |
| unet_block = ResUnetSkipConnectionBlock(ngf * 8, ngf * 8, input_nc=None, submodule=unet_block, | |
| norm_layer=norm_layer, use_dropout=use_dropout) | |
| unet_block = ResUnetSkipConnectionBlock(ngf * 4, ngf * 8, input_nc=None, submodule=unet_block, | |
| norm_layer=norm_layer) | |
| unet_block = ResUnetSkipConnectionBlock(ngf * 2, ngf * 4, input_nc=None, submodule=unet_block, | |
| norm_layer=norm_layer) | |
| unet_block = ResUnetSkipConnectionBlock(ngf, ngf * 2, input_nc=None, submodule=unet_block, | |
| norm_layer=norm_layer) | |
| unet_block = ResUnetSkipConnectionBlock(output_nc, ngf, input_nc=input_nc, submodule=unet_block, outermost=True, | |
| norm_layer=norm_layer) | |
| self.model = unet_block | |
| def forward(self, input): | |
| output = self.model(input) | |
| return output | |
| # Defines the submodule with skip connection. | |
| # X -------------------identity---------------------- X | |
| # |-- downsampling -- |submodule| -- upsampling --| | |
| class ResUnetSkipConnectionBlock(nn.Module): | |
| """ | |
| Unet Layers with Residual Connection | |
| """ | |
| def __init__(self, outer_nc, inner_nc, input_nc=None, | |
| submodule=None, outermost=False, innermost=False, norm_layer=nn.BatchNorm2d, use_dropout=False): | |
| super(ResUnetSkipConnectionBlock, self).__init__() | |
| self.outermost = outermost | |
| use_bias = norm_layer == nn.InstanceNorm2d | |
| if input_nc is None: | |
| input_nc = outer_nc | |
| downconv = nn.Conv2d(input_nc, inner_nc, kernel_size=3, | |
| stride=2, padding=1, bias=use_bias) | |
| # add two resblock | |
| res_downconv = [ResidualBlock(inner_nc, norm_layer), ResidualBlock(inner_nc, norm_layer)] | |
| res_upconv = [ResidualBlock(outer_nc, norm_layer), ResidualBlock(outer_nc, norm_layer)] | |
| # res_downconv = [ResidualBlock(inner_nc)] | |
| # res_upconv = [ResidualBlock(outer_nc)] | |
| downrelu = nn.ReLU(True) | |
| uprelu = nn.ReLU(True) | |
| if norm_layer != None: | |
| downnorm = norm_layer(inner_nc) | |
| upnorm = norm_layer(outer_nc) | |
| if outermost: | |
| upsample = nn.Upsample(scale_factor=2, mode='nearest') | |
| upconv = nn.Conv2d(inner_nc * 2, outer_nc, kernel_size=3, stride=1, padding=1, bias=use_bias) | |
| down = [downconv, downrelu] + res_downconv | |
| # up = [uprelu, upsample, upconv, upnorm] | |
| up = [upsample, upconv] | |
| model = down + [submodule] + up | |
| elif innermost: | |
| upsample = nn.Upsample(scale_factor=2, mode='nearest') | |
| upconv = nn.Conv2d(inner_nc, outer_nc, kernel_size=3, stride=1, padding=1, bias=use_bias) | |
| down = [downconv, downrelu] + res_downconv | |
| if norm_layer == None: | |
| up = [upsample, upconv, uprelu] + res_upconv | |
| else: | |
| up = [upsample, upconv, upnorm, uprelu] + res_upconv | |
| model = down + up | |
| else: | |
| upsample = nn.Upsample(scale_factor=2, mode='nearest') | |
| upconv = nn.Conv2d(inner_nc * 2, outer_nc, kernel_size=3, stride=1, padding=1, bias=use_bias) | |
| if norm_layer == None: | |
| down = [downconv, downrelu] + res_downconv | |
| up = [upsample, upconv, uprelu] + res_upconv | |
| else: | |
| down = [downconv, downnorm, downrelu] + res_downconv | |
| up = [upsample, upconv, upnorm, uprelu] + res_upconv | |
| if use_dropout: | |
| model = down + [submodule] + up + [nn.Dropout(0.5)] | |
| else: | |
| model = down + [submodule] + up | |
| self.model = nn.Sequential(*model) | |
| def forward(self, x): | |
| if self.outermost: | |
| return self.model(x) | |
| else: | |
| return torch.cat([x, self.model(x)], 1) | |
| class ResidualBlock(nn.Module): | |
| """ | |
| Residual Connection Layers | |
| """ | |
| def __init__(self, in_features=64, norm_layer=nn.BatchNorm2d): | |
| super(ResidualBlock, self).__init__() | |
| self.relu = nn.ReLU(True) | |
| if norm_layer == None: | |
| # hard to converge with out batch or instance norm | |
| self.block = nn.Sequential( | |
| nn.Conv2d(in_features, in_features, 3, 1, 1, bias=False), | |
| nn.ReLU(inplace=True), | |
| nn.Conv2d(in_features, in_features, 3, 1, 1, bias=False), | |
| ) | |
| else: | |
| self.block = nn.Sequential( | |
| nn.Conv2d(in_features, in_features, 3, 1, 1, bias=False), | |
| norm_layer(in_features), | |
| nn.ReLU(inplace=True), | |
| nn.Conv2d(in_features, in_features, 3, 1, 1, bias=False), | |
| norm_layer(in_features) | |
| ) | |
| def forward(self, x): | |
| residual = x | |
| out = self.block(x) | |
| out += residual | |
| out = self.relu(out) | |
| return out | |