# ------------------------------------------------------------------------ # Copyright (c) 2022 megvii-model. All Rights Reserved. # ------------------------------------------------------------------------ ''' Simple Baselines for Image Restoration @article{chen2022simple, title={Simple Baselines for Image Restoration}, author={Chen, Liangyu and Chu, Xiaojie and Zhang, Xiangyu and Sun, Jian}, journal={arXiv preprint arXiv:2204.04676}, year={2022} } ''' import torch import torch.nn as nn import torch.nn.functional as F # from .models.archs.arch_util import LayerNorm2d import sys sys.path.append('/ghome/zhuyr/Deref_RW/networks/') class LayerNormFunction(torch.autograd.Function): @staticmethod def forward(ctx, x, weight, bias, eps): ctx.eps = eps N, C, H, W = x.size() mu = x.mean(1, keepdim=True) var = (x - mu).pow(2).mean(1, keepdim=True) y = (x - mu) / (var + eps).sqrt() ctx.save_for_backward(y, var, weight) y = weight.view(1, C, 1, 1) * y + bias.view(1, C, 1, 1) return y @staticmethod def backward(ctx, grad_output): eps = ctx.eps N, C, H, W = grad_output.size() y, var, weight = ctx.saved_variables g = grad_output * weight.view(1, C, 1, 1) mean_g = g.mean(dim=1, keepdim=True) mean_gy = (g * y).mean(dim=1, keepdim=True) gx = 1. / torch.sqrt(var + eps) * (g - y * mean_gy - mean_g) return gx, (grad_output * y).sum(dim=3).sum(dim=2).sum(dim=0), grad_output.sum(dim=3).sum(dim=2).sum( dim=0), None class LayerNorm2d(nn.Module): def __init__(self, channels, eps=1e-6): super(LayerNorm2d, self).__init__() self.register_parameter('weight', nn.Parameter(torch.ones(channels))) self.register_parameter('bias', nn.Parameter(torch.zeros(channels))) self.eps = eps def forward(self, x): return LayerNormFunction.apply(x, self.weight, self.bias, self.eps) class SimpleGate(nn.Module): def forward(self, x): x1, x2 = x.chunk(2, dim=1) return x1 * x2 class NAFBlock(nn.Module): def __init__(self, c, DW_Expand=2, FFN_Expand=2, drop_out_rate=0.): super().__init__() dw_channel = c * DW_Expand self.conv1 = nn.Conv2d(in_channels=c, out_channels=dw_channel, kernel_size=1, padding=0, stride=1, groups=1, bias=True) self.conv2 = nn.Conv2d(in_channels=dw_channel, out_channels=dw_channel, kernel_size=3, padding=1, stride=1, groups=dw_channel, bias=True) self.conv3 = nn.Conv2d(in_channels=dw_channel // 2, out_channels=c, kernel_size=1, padding=0, stride=1, groups=1, bias=True) # Simplified Channel Attention self.sca = nn.Sequential( nn.AdaptiveAvgPool2d(1), nn.Conv2d(in_channels=dw_channel // 2, out_channels=dw_channel // 2, kernel_size=1, padding=0, stride=1, groups=1, bias=True), ) # SimpleGate self.sg = SimpleGate() ffn_channel = FFN_Expand * c self.conv4 = nn.Conv2d(in_channels=c, out_channels=ffn_channel, kernel_size=1, padding=0, stride=1, groups=1, bias=True) self.conv5 = nn.Conv2d(in_channels=ffn_channel // 2, out_channels=c, kernel_size=1, padding=0, stride=1, groups=1, bias=True) self.norm1 = LayerNorm2d(c) self.norm2 = LayerNorm2d(c) self.dropout1 = nn.Dropout(drop_out_rate) if drop_out_rate > 0. else nn.Identity() self.dropout2 = nn.Dropout(drop_out_rate) if drop_out_rate > 0. else nn.Identity() self.beta = nn.Parameter(torch.zeros((1, c, 1, 1)), requires_grad=True) self.gamma = nn.Parameter(torch.zeros((1, c, 1, 1)), requires_grad=True) def forward(self, inp): x = inp x = self.norm1(x) x = self.conv1(x) x = self.conv2(x) x = self.sg(x) x = x * self.sca(x) x = self.conv3(x) x = self.dropout1(x) y = inp + x * self.beta x = self.conv4(self.norm2(y)) x = self.sg(x) x = self.conv5(x) x = self.dropout2(x) return y + x * self.gamma class NAFNet(nn.Module): def __init__(self, img_channel=3, width=32, middle_blk_num=1, enc_blk_nums=[1, 1, 1, 28], dec_blk_nums=[1, 1, 1, 1], global_residual = False, drop_flag = False, drop_rate = 0.4): super().__init__() self.intro = nn.Conv2d(in_channels=img_channel, out_channels=width, kernel_size=3, padding=1, stride=1, groups=1, bias=True) self.ending = nn.Conv2d(in_channels=width, out_channels=3, kernel_size=3, padding=1, stride=1, groups=1, bias=True) self.encoders = nn.ModuleList() self.decoders = nn.ModuleList() self.middle_blks = nn.ModuleList() self.ups = nn.ModuleList() self.downs = nn.ModuleList() self.global_residual = global_residual self.drop_flag = drop_flag if drop_flag: self.dropout = nn.Dropout2d(p=drop_rate) chan = width for num in enc_blk_nums: self.encoders.append( nn.Sequential( *[NAFBlock(chan) for _ in range(num)] ) ) self.downs.append( nn.Conv2d(chan, 2*chan, 2, 2) ) chan = chan * 2 self.middle_blks = \ nn.Sequential( *[NAFBlock(chan) for _ in range(middle_blk_num)] ) for num in dec_blk_nums: self.ups.append( nn.Sequential( nn.Conv2d(chan, chan * 2, 1, bias=False), nn.PixelShuffle(2) ) ) chan = chan // 2 self.decoders.append( nn.Sequential( *[NAFBlock(chan) for _ in range(num)] ) ) self.padder_size = 2 ** len(self.encoders) def forward(self, inp): B, C, H, W = inp.shape inp = self.check_image_size(inp) base_inp = inp[:, :3, :, :] x = self.intro(inp) encs = [] for encoder, down in zip(self.encoders, self.downs): x = encoder(x) encs.append(x) x = down(x) x = self.middle_blks(x) for decoder, up, enc_skip in zip(self.decoders, self.ups, encs[::-1]): x = up(x) x = x + enc_skip x = decoder(x) if self.drop_flag: x = self.dropout(x) x = self.ending(x) if self.global_residual: #print(x.shape, inp.shape, base_inp.shape) x = x + base_inp else: x return x[:, :, :H, :W] def check_image_size(self, x): _, _, h, w = x.size() mod_pad_h = (self.padder_size - h % self.padder_size) % self.padder_size mod_pad_w = (self.padder_size - w % self.padder_size) % self.padder_size x = F.pad(x, (0, mod_pad_w, 0, mod_pad_h)) return x class NAFNet_wDetHead(nn.Module): def __init__(self, img_channel=3, width=32, middle_blk_num=1, enc_blk_nums=[1, 1, 1, 28], dec_blk_nums=[1, 1, 1, 1], global_residual = False, drop_flag = False, drop_rate = 0.4, concat = False, merge_manner = 0): super().__init__() self.intro = nn.Conv2d(in_channels=img_channel, out_channels=width, kernel_size=3, padding=1, stride=1, groups=1, bias=True) self.ending = nn.Conv2d(in_channels=width, out_channels=3, kernel_size=3, padding=1, stride=1, groups=1, bias=True) self.encoders = nn.ModuleList() self.decoders = nn.ModuleList() self.middle_blks = nn.ModuleList() self.ups = nn.ModuleList() self.downs = nn.ModuleList() self.global_residual = global_residual self.drop_flag = drop_flag self.concat = concat self.merge_manner = merge_manner if drop_flag: self.dropout = nn.Dropout2d(p=drop_rate) # --------------------------- Merge sparse & Img ------------------------------------------------------- self.intro_Det = nn.Conv2d(in_channels=1, out_channels=width, kernel_size=3, padding=1, stride=1, groups=1, bias=True) self.DetEnc = nn.Sequential( *[NAFBlock(width) for _ in range(3)] ) if self.concat: self.Merge_conv = nn.Conv2d(in_channels=width *2 , out_channels=width, kernel_size=3, padding=1, stride=1, groups=1, bias=True) else: self.Merge_conv = nn.Conv2d(in_channels=width , out_channels=width, kernel_size=3, padding=1, stride=1, groups=1, bias=True) # --------------------------- Merge sparse & Img ------------------------------------------------------- chan = width for num in enc_blk_nums: self.encoders.append( nn.Sequential( *[NAFBlock(chan) for _ in range(num)] ) ) self.downs.append( nn.Conv2d(chan, 2*chan, 2, 2) ) chan = chan * 2 self.middle_blks = \ nn.Sequential( *[NAFBlock(chan) for _ in range(middle_blk_num)] ) for num in dec_blk_nums: self.ups.append( nn.Sequential( nn.Conv2d(chan, chan * 2, 1, bias=False), nn.PixelShuffle(2) ) ) chan = chan // 2 self.decoders.append( nn.Sequential( *[NAFBlock(chan) for _ in range(num)] ) ) self.padder_size = 2 ** len(self.encoders) def forward(self, inp, spare_ref): B, C, H, W = inp.shape inp = self.check_image_size(inp) base_inp = inp #[:, :3, :, :] x = self.intro(inp) fea_sparse = self.DetEnc(self.intro_Det(spare_ref)) if self.merge_manner ==0 and self.concat: x = torch.cat([x, fea_sparse], dim=1) x = self.Merge_conv(x) elif self.merge_manner == 1 and not self.concat: x = x + fea_sparse x = self.Merge_conv(x) elif self.merge_manner == 2 and not self.concat: x = x + fea_sparse *x x = self.Merge_conv(x) else: x = x print('Merge Flag Error!!!(No Merge Operation) ---zyr 1031 ') encs = [] for encoder, down in zip(self.encoders, self.downs): x = encoder(x) encs.append(x) x = down(x) x = self.middle_blks(x) for decoder, up, enc_skip in zip(self.decoders, self.ups, encs[::-1]): x = up(x) x = x + enc_skip x = decoder(x) if self.drop_flag: x = self.dropout(x) x = self.ending(x) if self.global_residual: #print(x.shape, inp.shape, base_inp.shape) x = x + base_inp else: x return x[:, :, :H, :W] def check_image_size(self, x): _, _, h, w = x.size() mod_pad_h = (self.padder_size - h % self.padder_size) % self.padder_size mod_pad_w = (self.padder_size - w % self.padder_size) % self.padder_size x = F.pad(x, (0, mod_pad_w, 0, mod_pad_h)) return x class NAFNet_refine(nn.Module): def __init__(self, img_channel=6, width=32, middle_blk_num=1, enc_blk_nums=[1, 1, 1, 28], dec_blk_nums=[1, 1, 1, 1], global_residual = False): super().__init__() self.intro = nn.Conv2d(in_channels=img_channel, out_channels=width, kernel_size=3, padding=1, stride=1, groups=1, bias=True) self.ending = nn.Conv2d(in_channels=width, out_channels=3, kernel_size=3, padding=1, stride=1, groups=1, bias=True) self.encoders = nn.ModuleList() self.decoders = nn.ModuleList() self.middle_blks = nn.ModuleList() self.ups = nn.ModuleList() self.downs = nn.ModuleList() self.global_residual = global_residual chan = width for num in enc_blk_nums: self.encoders.append( nn.Sequential( *[NAFBlock(chan) for _ in range(num)] ) ) self.downs.append( nn.Conv2d(chan, 2*chan, 2, 2) ) chan = chan * 2 self.middle_blks = \ nn.Sequential( *[NAFBlock(chan) for _ in range(middle_blk_num)] ) for num in dec_blk_nums: self.ups.append( nn.Sequential( nn.Conv2d(chan, chan * 2, 1, bias=False), nn.PixelShuffle(2) ) ) chan = chan // 2 self.decoders.append( nn.Sequential( *[NAFBlock(chan) for _ in range(num)] ) ) self.padder_size = 2 ** len(self.encoders) def forward(self, inp, pre_pred): B, C, H, W = inp.shape inp = self.check_image_size(inp) pre_pred = self.check_image_size(pre_pred) network_in = torch.cat([inp, pre_pred ], dim= 1) x = self.intro(network_in) encs = [] for encoder, down in zip(self.encoders, self.downs): x = encoder(x) encs.append(x) x = down(x) x = self.middle_blks(x) for decoder, up, enc_skip in zip(self.decoders, self.ups, encs[::-1]): x = up(x) x = x + enc_skip x = decoder(x) x = self.ending(x) if self.global_residual: x = x + inp[:3,:,:,:] else: x return x[:, :, :H, :W] def check_image_size(self, x): _, _, h, w = x.size() mod_pad_h = (self.padder_size - h % self.padder_size) % self.padder_size mod_pad_w = (self.padder_size - w % self.padder_size) % self.padder_size x = F.pad(x, (0, mod_pad_w, 0, mod_pad_h)) return x def print_param_number(net): print('#generator parameters:', sum(param.numel() for param in net.parameters())) if __name__ == '__main__': img_channel = 3 width = 32 # enc_blks = [2, 2, 4, 8] # middle_blk_num = 12 # dec_blks = [2, 2, 2, 2] # enc_blks = [2, 2, 4, 8] # middle_blk_num = 12 # dec_blks = [2, 2, 2, 2] # enc_blks = [1, 1, 1, 28] # middle_blk_num = 1 # dec_blks = [1, 1, 1, 1] enc_blks = [1, 1, 1, 28] middle_blk_num = 1 dec_blks = [1, 1, 1, 1] net = NAFNet_wDetHead(img_channel=img_channel, width=width, middle_blk_num=middle_blk_num, enc_blk_nums=enc_blks, dec_blk_nums=dec_blks,global_residual = True, concat= True, merge_manner= 2) #.cuda() #print(net) size = 352 input = torch.randn([1,3,128, 128])#.cuda() inp_shape = (5, 3, 128, 128) spare = torch.randn([1,1,128, 128]) print(net(input, spare).size()) print_param_number(net) #net_local = NAFNetLocal()#.cuda() #print_param_number(net) # print(net_local(input).size()) # inp_shape = (3, 256, 256) # # from ptflops import get_model_complexity_info # # macs, params = get_model_complexity_info(net, inp_shape, verbose=False, print_per_layer_stat=False) # # params = float(params[:-3]) # macs = float(macs[:-4]) # # print(macs, params)