""" ## Uformer: A General U-Shaped Transformer for Image Restoration ## Zhendong Wang, Xiaodong Cun, Jianmin Bao, Jianzhuang Liu ## https://arxiv.org/abs/2106.03106 """ import torch import torch.nn as nn import torch.utils.checkpoint as checkpoint from timm.models.layers import DropPath, to_2tuple, trunc_normal_ import torch.nn.functional as F from einops import rearrange, repeat from einops.layers.torch import Rearrange import math import numpy as np import time from torch import einsum # from basicsr.utils.registry import ARCH_REGISTRY ######################################### class ConvBlock(nn.Module): def __init__(self, in_channel, out_channel, strides=1): super(ConvBlock, self).__init__() self.strides = strides self.in_channel = in_channel self.out_channel = out_channel self.block = nn.Sequential( nn.Conv2d( in_channel, out_channel, kernel_size=3, stride=strides, padding=1 ), nn.LeakyReLU(inplace=True), nn.Conv2d( out_channel, out_channel, kernel_size=3, stride=strides, padding=1 ), nn.LeakyReLU(inplace=True), ) self.conv11 = nn.Conv2d( in_channel, out_channel, kernel_size=1, stride=strides, padding=0 ) def forward(self, x): out1 = self.block(x) out2 = self.conv11(x) out = out1 + out2 return out def flops(self, H, W): flops = ( H * W * self.in_channel * self.out_channel * (3 * 3 + 1) + H * W * self.out_channel * self.out_channel * 3 * 3 ) return flops class UNet(nn.Module): def __init__(self, block=ConvBlock, dim=32): super(UNet, self).__init__() self.dim = dim self.ConvBlock1 = ConvBlock(3, dim, strides=1) self.pool1 = nn.Conv2d(dim, dim, kernel_size=4, stride=2, padding=1) self.ConvBlock2 = block(dim, dim * 2, strides=1) self.pool2 = nn.Conv2d(dim * 2, dim * 2, kernel_size=4, stride=2, padding=1) self.ConvBlock3 = block(dim * 2, dim * 4, strides=1) self.pool3 = nn.Conv2d(dim * 4, dim * 4, kernel_size=4, stride=2, padding=1) self.ConvBlock4 = block(dim * 4, dim * 8, strides=1) self.pool4 = nn.Conv2d(dim * 8, dim * 8, kernel_size=4, stride=2, padding=1) self.ConvBlock5 = block(dim * 8, dim * 16, strides=1) self.upv6 = nn.ConvTranspose2d(dim * 16, dim * 8, 2, stride=2) self.ConvBlock6 = block(dim * 16, dim * 8, strides=1) self.upv7 = nn.ConvTranspose2d(dim * 8, dim * 4, 2, stride=2) self.ConvBlock7 = block(dim * 8, dim * 4, strides=1) self.upv8 = nn.ConvTranspose2d(dim * 4, dim * 2, 2, stride=2) self.ConvBlock8 = block(dim * 4, dim * 2, strides=1) self.upv9 = nn.ConvTranspose2d(dim * 2, dim, 2, stride=2) self.ConvBlock9 = block(dim * 2, dim, strides=1) self.conv10 = nn.Conv2d(dim, 3, kernel_size=3, stride=1, padding=1) def forward(self, x): conv1 = self.ConvBlock1(x) pool1 = self.pool1(conv1) conv2 = self.ConvBlock2(pool1) pool2 = self.pool2(conv2) conv3 = self.ConvBlock3(pool2) pool3 = self.pool3(conv3) conv4 = self.ConvBlock4(pool3) pool4 = self.pool4(conv4) conv5 = self.ConvBlock5(pool4) up6 = self.upv6(conv5) up6 = torch.cat([up6, conv4], 1) conv6 = self.ConvBlock6(up6) up7 = self.upv7(conv6) up7 = torch.cat([up7, conv3], 1) conv7 = self.ConvBlock7(up7) up8 = self.upv8(conv7) up8 = torch.cat([up8, conv2], 1) conv8 = self.ConvBlock8(up8) up9 = self.upv9(conv8) up9 = torch.cat([up9, conv1], 1) conv9 = self.ConvBlock9(up9) conv10 = self.conv10(conv9) out = x + conv10 return out def flops(self, H, W): flops = 0 flops += self.ConvBlock1.flops(H, W) flops += H / 2 * W / 2 * self.dim * self.dim * 4 * 4 flops += self.ConvBlock2.flops(H / 2, W / 2) flops += H / 4 * W / 4 * self.dim * 2 * self.dim * 2 * 4 * 4 flops += self.ConvBlock3.flops(H / 4, W / 4) flops += H / 8 * W / 8 * self.dim * 4 * self.dim * 4 * 4 * 4 flops += self.ConvBlock4.flops(H / 8, W / 8) flops += H / 16 * W / 16 * self.dim * 8 * self.dim * 8 * 4 * 4 flops += self.ConvBlock5.flops(H / 16, W / 16) flops += H / 8 * W / 8 * self.dim * 16 * self.dim * 8 * 2 * 2 flops += self.ConvBlock6.flops(H / 8, W / 8) flops += H / 4 * W / 4 * self.dim * 8 * self.dim * 4 * 2 * 2 flops += self.ConvBlock7.flops(H / 4, W / 4) flops += H / 2 * W / 2 * self.dim * 4 * self.dim * 2 * 2 * 2 flops += self.ConvBlock8.flops(H / 2, W / 2) flops += H * W * self.dim * 2 * self.dim * 2 * 2 flops += self.ConvBlock9.flops(H, W) flops += H * W * self.dim * 3 * 3 * 3 return flops ######################################### class PosCNN(nn.Module): def __init__(self, in_chans, embed_dim=768, s=1): super(PosCNN, self).__init__() self.proj = nn.Sequential( nn.Conv2d(in_chans, embed_dim, 3, s, 1, bias=True, groups=embed_dim) ) self.s = s def forward(self, x, H=None, W=None): B, N, C = x.shape H = H or int(math.sqrt(N)) W = W or int(math.sqrt(N)) feat_token = x cnn_feat = feat_token.transpose(1, 2).view(B, C, H, W) if self.s == 1: x = self.proj(cnn_feat) + cnn_feat else: x = self.proj(cnn_feat) x = x.flatten(2).transpose(1, 2) return x def no_weight_decay(self): return ["proj.%d.weight" % i for i in range(4)] class SELayer(nn.Module): def __init__(self, channel, reduction=16): super(SELayer, self).__init__() self.avg_pool = nn.AdaptiveAvgPool1d(1) self.fc = nn.Sequential( nn.Linear(channel, channel // reduction, bias=False), nn.ReLU(inplace=True), nn.Linear(channel // reduction, channel, bias=False), nn.Sigmoid(), ) def forward(self, x): # x: [B, N, C] x = torch.transpose(x, 1, 2) # [B, C, N] b, c, _ = x.size() y = self.avg_pool(x).view(b, c) y = self.fc(y).view(b, c, 1) x = x * y.expand_as(x) x = torch.transpose(x, 1, 2) # [B, N, C] return x class SepConv2d(torch.nn.Module): def __init__( self, in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, act_layer=nn.ReLU, ): super(SepConv2d, self).__init__() self.depthwise = torch.nn.Conv2d( in_channels, in_channels, kernel_size=kernel_size, stride=stride, padding=padding, dilation=dilation, groups=in_channels, ) self.pointwise = torch.nn.Conv2d(in_channels, out_channels, kernel_size=1) self.act_layer = act_layer() if act_layer is not None else nn.Identity() self.in_channels = in_channels self.out_channels = out_channels self.kernel_size = kernel_size self.stride = stride def forward(self, x): x = self.depthwise(x) x = self.act_layer(x) x = self.pointwise(x) return x def flops(self, H, W): flops = 0 flops += H * W * self.in_channels * self.kernel_size**2 / self.stride**2 flops += H * W * self.in_channels * self.out_channels return flops ######################################### ######## Embedding for q,k,v ######## class ConvProjection(nn.Module): def __init__( self, dim, heads=8, dim_head=64, kernel_size=3, q_stride=1, k_stride=1, v_stride=1, dropout=0.0, last_stage=False, bias=True, ): super().__init__() inner_dim = dim_head * heads self.heads = heads pad = (kernel_size - q_stride) // 2 self.to_q = SepConv2d(dim, inner_dim, kernel_size, q_stride, pad, bias) self.to_k = SepConv2d(dim, inner_dim, kernel_size, k_stride, pad, bias) self.to_v = SepConv2d(dim, inner_dim, kernel_size, v_stride, pad, bias) def forward(self, x, attn_kv=None): b, n, c, h = *x.shape, self.heads l = int(math.sqrt(n)) w = int(math.sqrt(n)) attn_kv = x if attn_kv is None else attn_kv x = rearrange(x, "b (l w) c -> b c l w", l=l, w=w) attn_kv = rearrange(attn_kv, "b (l w) c -> b c l w", l=l, w=w) # print(attn_kv) q = self.to_q(x) q = rearrange(q, "b (h d) l w -> b h (l w) d", h=h) k = self.to_k(attn_kv) v = self.to_v(attn_kv) k = rearrange(k, "b (h d) l w -> b h (l w) d", h=h) v = rearrange(v, "b (h d) l w -> b h (l w) d", h=h) return q, k, v def flops(self, H, W): flops = 0 flops += self.to_q.flops(H, W) flops += self.to_k.flops(H, W) flops += self.to_v.flops(H, W) return flops class LinearProjection(nn.Module): def __init__(self, dim, heads=8, dim_head=64, dropout=0.0, bias=True): super().__init__() inner_dim = dim_head * heads self.heads = heads self.to_q = nn.Linear(dim, inner_dim, bias=bias) self.to_kv = nn.Linear(dim, inner_dim * 2, bias=bias) self.dim = dim self.inner_dim = inner_dim def forward(self, x, attn_kv=None): B_, N, C = x.shape attn_kv = x if attn_kv is None else attn_kv q = ( self.to_q(x) .reshape(B_, N, 1, self.heads, C // self.heads) .permute(2, 0, 3, 1, 4) ) kv = ( self.to_kv(attn_kv) .reshape(B_, N, 2, self.heads, C // self.heads) .permute(2, 0, 3, 1, 4) ) q = q[0] k, v = kv[0], kv[1] return q, k, v def flops(self, H, W): flops = H * W * self.dim * self.inner_dim * 3 return flops class LinearProjection_Concat_kv(nn.Module): def __init__(self, dim, heads=8, dim_head=64, dropout=0.0, bias=True): super().__init__() inner_dim = dim_head * heads self.heads = heads self.to_qkv = nn.Linear(dim, inner_dim * 3, bias=bias) self.to_kv = nn.Linear(dim, inner_dim * 2, bias=bias) self.dim = dim self.inner_dim = inner_dim def forward(self, x, attn_kv=None): B_, N, C = x.shape attn_kv = x if attn_kv is None else attn_kv qkv_dec = ( self.to_qkv(x) .reshape(B_, N, 3, self.heads, C // self.heads) .permute(2, 0, 3, 1, 4) ) kv_enc = ( self.to_kv(attn_kv) .reshape(B_, N, 2, self.heads, C // self.heads) .permute(2, 0, 3, 1, 4) ) q, k_d, v_d = ( qkv_dec[0], qkv_dec[1], qkv_dec[2], ) # make torchscript happy (cannot use tensor as tuple) k_e, v_e = kv_enc[0], kv_enc[1] k = torch.cat((k_d, k_e), dim=2) v = torch.cat((v_d, v_e), dim=2) return q, k, v def flops(self, H, W): flops = H * W * self.dim * self.inner_dim * 5 return flops ######################################### ########### window-based self-attention ############# class WindowAttention(nn.Module): def __init__( self, dim, win_size, num_heads, token_projection="linear", qkv_bias=True, qk_scale=None, attn_drop=0.0, proj_drop=0.0, se_layer=False, ): super().__init__() self.dim = dim self.win_size = win_size # Wh, Ww self.num_heads = num_heads head_dim = dim // num_heads self.scale = qk_scale or head_dim**-0.5 # define a parameter table of relative position bias self.relative_position_bias_table = nn.Parameter( torch.zeros((2 * win_size[0] - 1) * (2 * win_size[1] - 1), num_heads) ) # 2*Wh-1 * 2*Ww-1, nH # get pair-wise relative position index for each token inside the window coords_h = torch.arange(self.win_size[0]) # [0,...,Wh-1] coords_w = torch.arange(self.win_size[1]) # [0,...,Ww-1] coords = torch.stack(torch.meshgrid([coords_h, coords_w])) # 2, Wh, Ww coords_flatten = torch.flatten(coords, 1) # 2, Wh*Ww relative_coords = ( coords_flatten[:, :, None] - coords_flatten[:, None, :] ) # 2, Wh*Ww, Wh*Ww relative_coords = relative_coords.permute( 1, 2, 0 ).contiguous() # Wh*Ww, Wh*Ww, 2 relative_coords[:, :, 0] += self.win_size[0] - 1 # shift to start from 0 relative_coords[:, :, 1] += self.win_size[1] - 1 relative_coords[:, :, 0] *= 2 * self.win_size[1] - 1 relative_position_index = relative_coords.sum(-1) # Wh*Ww, Wh*Ww self.register_buffer("relative_position_index", relative_position_index) # self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias) if token_projection == "conv": self.qkv = ConvProjection(dim, num_heads, dim // num_heads, bias=qkv_bias) elif token_projection == "linear_concat": self.qkv = LinearProjection_Concat_kv( dim, num_heads, dim // num_heads, bias=qkv_bias ) else: self.qkv = LinearProjection(dim, num_heads, dim // num_heads, bias=qkv_bias) self.token_projection = token_projection self.attn_drop = nn.Dropout(attn_drop) self.proj = nn.Linear(dim, dim) self.se_layer = SELayer(dim) if se_layer else nn.Identity() self.proj_drop = nn.Dropout(proj_drop) trunc_normal_(self.relative_position_bias_table, std=0.02) self.softmax = nn.Softmax(dim=-1) def forward(self, x, attn_kv=None, mask=None): B_, N, C = x.shape q, k, v = self.qkv(x, attn_kv) q = q * self.scale attn = q @ k.transpose(-2, -1) relative_position_bias = self.relative_position_bias_table[ self.relative_position_index.view(-1) ].view( self.win_size[0] * self.win_size[1], self.win_size[0] * self.win_size[1], -1 ) # Wh*Ww,Wh*Ww,nH relative_position_bias = relative_position_bias.permute( 2, 0, 1 ).contiguous() # nH, Wh*Ww, Wh*Ww ratio = attn.size(-1) // relative_position_bias.size(-1) relative_position_bias = repeat( relative_position_bias, "nH l c -> nH l (c d)", d=ratio ) attn = attn + relative_position_bias.unsqueeze(0) if mask is not None: nW = mask.shape[0] mask = repeat(mask, "nW m n -> nW m (n d)", d=ratio) attn = attn.view( B_ // nW, nW, self.num_heads, N, N * ratio ) + mask.unsqueeze(1).unsqueeze(0) attn = attn.view(-1, self.num_heads, N, N * ratio) attn = self.softmax(attn) else: attn = self.softmax(attn) attn = self.attn_drop(attn) x = (attn @ v).transpose(1, 2).reshape(B_, N, C) x = self.proj(x) x = self.se_layer(x) x = self.proj_drop(x) return x def extra_repr(self) -> str: return f"dim={self.dim}, win_size={self.win_size}, num_heads={self.num_heads}" def flops(self, H, W): # calculate flops for 1 window with token length of N # print(N, self.dim) flops = 0 N = self.win_size[0] * self.win_size[1] nW = H * W / N # qkv = self.qkv(x) # flops += N * self.dim * 3 * self.dim flops += self.qkv.flops(H, W) # attn = (q @ k.transpose(-2, -1)) if self.token_projection != "linear_concat": flops += nW * self.num_heads * N * (self.dim // self.num_heads) * N # x = (attn @ v) flops += nW * self.num_heads * N * N * (self.dim // self.num_heads) else: flops += nW * self.num_heads * N * (self.dim // self.num_heads) * N * 2 # x = (attn @ v) flops += nW * self.num_heads * N * N * 2 * (self.dim // self.num_heads) # x = self.proj(x) flops += nW * N * self.dim * self.dim print("W-MSA:{%.2f}" % (flops / 1e9)) return flops ######################################### ########### feed-forward network ############# class Mlp(nn.Module): def __init__( self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.0, ): super().__init__() out_features = out_features or in_features hidden_features = hidden_features or in_features self.fc1 = nn.Linear(in_features, hidden_features) self.act = act_layer() self.fc2 = nn.Linear(hidden_features, out_features) self.drop = nn.Dropout(drop) self.in_features = in_features self.hidden_features = hidden_features self.out_features = out_features def forward(self, x): x = self.fc1(x) x = self.act(x) x = self.drop(x) x = self.fc2(x) x = self.drop(x) return x def flops(self, H, W): flops = 0 # fc1 flops += H * W * self.in_features * self.hidden_features # fc2 flops += H * W * self.hidden_features * self.out_features print("MLP:{%.2f}" % (flops / 1e9)) return flops class LeFF(nn.Module): def __init__(self, dim=32, hidden_dim=128, act_layer=nn.GELU, drop=0.0): super().__init__() self.linear1 = nn.Sequential(nn.Linear(dim, hidden_dim), act_layer()) self.dwconv = nn.Sequential( nn.Conv2d( hidden_dim, hidden_dim, groups=hidden_dim, kernel_size=3, stride=1, padding=1, ), act_layer(), ) self.linear2 = nn.Sequential(nn.Linear(hidden_dim, dim)) self.dim = dim self.hidden_dim = hidden_dim def forward(self, x): # bs x hw x c bs, hw, c = x.size() hh = int(math.sqrt(hw)) x = self.linear1(x) # spatial restore x = rearrange(x, " b (h w) (c) -> b c h w ", h=hh, w=hh) # bs,hidden_dim,32x32 x = self.dwconv(x) # flaten x = rearrange(x, " b c h w -> b (h w) c", h=hh, w=hh) x = self.linear2(x) return x def flops(self, H, W): flops = 0 # fc1 flops += H * W * self.dim * self.hidden_dim # dwconv flops += H * W * self.hidden_dim * 3 * 3 # fc2 flops += H * W * self.hidden_dim * self.dim print("LeFF:{%.2f}" % (flops / 1e9)) return flops ######################################### ########### window operation############# def window_partition(x, win_size, dilation_rate=1): B, H, W, C = x.shape if dilation_rate != 1: x = x.permute(0, 3, 1, 2) # B, C, H, W assert type(dilation_rate) is int, "dilation_rate should be a int" x = F.unfold( x, kernel_size=win_size, dilation=dilation_rate, padding=4 * (dilation_rate - 1), stride=win_size, ) # B, C*Wh*Ww, H/Wh*W/Ww windows = ( x.permute(0, 2, 1).contiguous().view(-1, C, win_size, win_size) ) # B' ,C ,Wh ,Ww windows = windows.permute(0, 2, 3, 1).contiguous() # B' ,Wh ,Ww ,C else: x = x.view(B, H // win_size, win_size, W // win_size, win_size, C) windows = ( x.permute(0, 1, 3, 2, 4, 5).contiguous().view(-1, win_size, win_size, C) ) # B' ,Wh ,Ww ,C return windows def window_reverse(windows, win_size, H, W, dilation_rate=1): # B' ,Wh ,Ww ,C B = int(windows.shape[0] / (H * W / win_size / win_size)) x = windows.view(B, H // win_size, W // win_size, win_size, win_size, -1) if dilation_rate != 1: x = windows.permute(0, 5, 3, 4, 1, 2).contiguous() # B, C*Wh*Ww, H/Wh*W/Ww x = F.fold( x, (H, W), kernel_size=win_size, dilation=dilation_rate, padding=4 * (dilation_rate - 1), stride=win_size, ) else: x = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(B, H, W, -1) return x ######################################### # Downsample Block class Downsample(nn.Module): def __init__(self, in_channel, out_channel): super(Downsample, self).__init__() self.conv = nn.Sequential( nn.Conv2d(in_channel, out_channel, kernel_size=4, stride=2, padding=1), ) self.in_channel = in_channel self.out_channel = out_channel def forward(self, x): B, L, C = x.shape # import pdb;pdb.set_trace() H = int(math.sqrt(L)) W = int(math.sqrt(L)) x = x.transpose(1, 2).contiguous().view(B, C, H, W) out = self.conv(x).flatten(2).transpose(1, 2).contiguous() # B H*W C return out def flops(self, H, W): flops = 0 # conv flops += H / 2 * W / 2 * self.in_channel * self.out_channel * 4 * 4 print("Downsample:{%.2f}" % (flops / 1e9)) return flops # Upsample Block class Upsample(nn.Module): def __init__(self, in_channel, out_channel): super(Upsample, self).__init__() self.deconv = nn.Sequential( nn.ConvTranspose2d(in_channel, out_channel, kernel_size=2, stride=2), ) self.in_channel = in_channel self.out_channel = out_channel def forward(self, x): B, L, C = x.shape H = int(math.sqrt(L)) W = int(math.sqrt(L)) x = x.transpose(1, 2).contiguous().view(B, C, H, W) out = self.deconv(x).flatten(2).transpose(1, 2).contiguous() # B H*W C return out def flops(self, H, W): flops = 0 # conv flops += H * 2 * W * 2 * self.in_channel * self.out_channel * 2 * 2 print("Upsample:{%.2f}" % (flops / 1e9)) return flops # Input Projection class InputProj(nn.Module): def __init__( self, in_channel=3, out_channel=64, kernel_size=3, stride=1, norm_layer=None, act_layer=nn.LeakyReLU, ): super().__init__() self.proj = nn.Sequential( nn.Conv2d( in_channel, out_channel, kernel_size=3, stride=stride, padding=kernel_size // 2, ), act_layer(inplace=True), ) if norm_layer is not None: self.norm = norm_layer(out_channel) else: self.norm = None self.in_channel = in_channel self.out_channel = out_channel def forward(self, x): B, C, H, W = x.shape x = self.proj(x).flatten(2).transpose(1, 2).contiguous() # B H*W C if self.norm is not None: x = self.norm(x) return x def flops(self, H, W): flops = 0 # conv flops += H * W * self.in_channel * self.out_channel * 3 * 3 if self.norm is not None: flops += H * W * self.out_channel print("Input_proj:{%.2f}" % (flops / 1e9)) return flops # Output Projection class OutputProj(nn.Module): def __init__( self, in_channel=64, out_channel=3, kernel_size=3, stride=1, norm_layer=None, act_layer=None, ): super().__init__() self.proj = nn.Sequential( nn.Conv2d( in_channel, out_channel, kernel_size=3, stride=stride, padding=kernel_size // 2, ), ) if act_layer is not None: self.proj.add_module(act_layer(inplace=True)) if norm_layer is not None: self.norm = norm_layer(out_channel) else: self.norm = None self.in_channel = in_channel self.out_channel = out_channel def forward(self, x): B, L, C = x.shape H = int(math.sqrt(L)) W = int(math.sqrt(L)) x = x.transpose(1, 2).view(B, C, H, W) x = self.proj(x) if self.norm is not None: x = self.norm(x) return x def flops(self, H, W): flops = 0 # conv flops += H * W * self.in_channel * self.out_channel * 3 * 3 if self.norm is not None: flops += H * W * self.out_channel print("Output_proj:{%.2f}" % (flops / 1e9)) return flops ######################################### ########### LeWinTransformer ############# class LeWinTransformerBlock(nn.Module): def __init__( self, dim, input_resolution, num_heads, win_size=8, shift_size=0, mlp_ratio=4.0, qkv_bias=True, qk_scale=None, drop=0.0, attn_drop=0.0, drop_path=0.0, act_layer=nn.GELU, norm_layer=nn.LayerNorm, token_projection="linear", token_mlp="leff", se_layer=False, ): super().__init__() self.dim = dim self.input_resolution = input_resolution self.num_heads = num_heads self.win_size = win_size self.shift_size = shift_size self.mlp_ratio = mlp_ratio self.token_mlp = token_mlp if min(self.input_resolution) <= self.win_size: self.shift_size = 0 self.win_size = min(self.input_resolution) assert 0 <= self.shift_size < self.win_size, "shift_size must in 0-win_size" self.norm1 = norm_layer(dim) self.attn = WindowAttention( dim, win_size=to_2tuple(self.win_size), num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale, attn_drop=attn_drop, proj_drop=drop, token_projection=token_projection, se_layer=se_layer, ) self.drop_path = DropPath(drop_path) if drop_path > 0.0 else nn.Identity() self.norm2 = norm_layer(dim) mlp_hidden_dim = int(dim * mlp_ratio) self.mlp = ( Mlp( in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop, ) if token_mlp == "ffn" else LeFF(dim, mlp_hidden_dim, act_layer=act_layer, drop=drop) ) def extra_repr(self) -> str: return ( f"dim={self.dim}, input_resolution={self.input_resolution}, num_heads={self.num_heads}, " f"win_size={self.win_size}, shift_size={self.shift_size}, mlp_ratio={self.mlp_ratio}" ) def forward(self, x, mask=None): B, L, C = x.shape H = int(math.sqrt(L)) W = int(math.sqrt(L)) ## input mask if mask != None: input_mask = F.interpolate(mask, size=(H, W)).permute(0, 2, 3, 1) input_mask_windows = window_partition( input_mask, self.win_size ) # nW, win_size, win_size, 1 attn_mask = input_mask_windows.view( -1, self.win_size * self.win_size ) # nW, win_size*win_size attn_mask = attn_mask.unsqueeze(2) * attn_mask.unsqueeze( 1 ) # nW, win_size*win_size, win_size*win_size attn_mask = attn_mask.masked_fill( attn_mask != 0, float(-100.0) ).masked_fill(attn_mask == 0, float(0.0)) else: attn_mask = None ## shift mask if self.shift_size > 0: # calculate attention mask for SW-MSA shift_mask = torch.zeros((1, H, W, 1)).type_as(x) h_slices = ( slice(0, -self.win_size), slice(-self.win_size, -self.shift_size), slice(-self.shift_size, None), ) w_slices = ( slice(0, -self.win_size), slice(-self.win_size, -self.shift_size), slice(-self.shift_size, None), ) cnt = 0 for h in h_slices: for w in w_slices: shift_mask[:, h, w, :] = cnt cnt += 1 shift_mask_windows = window_partition( shift_mask, self.win_size ) # nW, win_size, win_size, 1 shift_mask_windows = shift_mask_windows.view( -1, self.win_size * self.win_size ) # nW, win_size*win_size shift_attn_mask = shift_mask_windows.unsqueeze( 1 ) - shift_mask_windows.unsqueeze( 2 ) # nW, win_size*win_size, win_size*win_size shift_attn_mask = shift_attn_mask.masked_fill( shift_attn_mask != 0, float(-100.0) ).masked_fill(shift_attn_mask == 0, float(0.0)) attn_mask = ( attn_mask + shift_attn_mask if attn_mask is not None else shift_attn_mask ) shortcut = x x = self.norm1(x) x = x.view(B, H, W, C) # cyclic shift if self.shift_size > 0: shifted_x = torch.roll( x, shifts=(-self.shift_size, -self.shift_size), dims=(1, 2) ) else: shifted_x = x # partition windows x_windows = window_partition( shifted_x, self.win_size ) # nW*B, win_size, win_size, C N*C->C x_windows = x_windows.view( -1, self.win_size * self.win_size, C ) # nW*B, win_size*win_size, C # W-MSA/SW-MSA attn_windows = self.attn( x_windows, mask=attn_mask ) # nW*B, win_size*win_size, C # merge windows attn_windows = attn_windows.view(-1, self.win_size, self.win_size, C) shifted_x = window_reverse(attn_windows, self.win_size, H, W) # B H' W' C # reverse cyclic shift if self.shift_size > 0: x = torch.roll( shifted_x, shifts=(self.shift_size, self.shift_size), dims=(1, 2) ) else: x = shifted_x x = x.view(B, H * W, C) # FFN x = shortcut + self.drop_path(x) x = x + self.drop_path(self.mlp(self.norm2(x))) del attn_mask return x def flops(self): flops = 0 H, W = self.input_resolution # norm1 flops += self.dim * H * W # W-MSA/SW-MSA flops += self.attn.flops(H, W) # norm2 flops += self.dim * H * W # mlp flops += self.mlp.flops(H, W) print("LeWin:{%.2f}" % (flops / 1e9)) return flops ########### LeWinTransformer_Cross ############# class LeWinTransformer_Cross(nn.Module): def __init__( self, dim, input_resolution, num_heads, win_size=8, shift_size=0, mlp_ratio=4.0, qkv_bias=True, qk_scale=None, drop=0.0, attn_drop=0.0, drop_path=0.0, act_layer=nn.GELU, norm_layer=nn.LayerNorm, token_projection="linear", token_mlp="ffn", ): super().__init__() self.dim = dim self.input_resolution = input_resolution self.num_heads = num_heads self.win_size = win_size self.shift_size = shift_size self.mlp_ratio = mlp_ratio if min(self.input_resolution) <= self.win_size: self.shift_size = 0 self.win_size = min(self.input_resolution) assert 0 <= self.shift_size < self.win_size, "shift_size must in 0-win_size" self.norm1 = norm_layer(dim) self.attn = WindowAttention( dim, win_size=to_2tuple(self.win_size), num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale, attn_drop=attn_drop, proj_drop=drop, token_projection=token_projection, ) self.norm2 = norm_layer(dim) self.norm_kv = norm_layer(dim) self.cross_attn = WindowAttention( dim, win_size=to_2tuple(self.win_size), num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale, attn_drop=attn_drop, proj_drop=drop, token_projection=token_projection, ) self.drop_path = DropPath(drop_path) if drop_path > 0.0 else nn.Identity() self.norm3 = norm_layer(dim) mlp_hidden_dim = int(dim * mlp_ratio) self.mlp = ( Mlp( in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop, ) if token_mlp == "ffn" else LeFF(dim, mlp_hidden_dim, act_layer=act_layer, drop=drop) ) def extra_repr(self) -> str: return ( f"dim={self.dim}, input_resolution={self.input_resolution}, num_heads={self.num_heads}, " f"win_size={self.win_size}, shift_size={self.shift_size}, mlp_ratio={self.mlp_ratio}" ) def forward(self, x, attn_kv=None, mask=None): B, L, C = x.shape H = int(math.sqrt(L)) W = int(math.sqrt(L)) ## input mask if mask != None: input_mask = F.interpolate(mask, size=(H, W)).permute(0, 2, 3, 1) input_mask_windows = window_partition( input_mask, self.win_size ) # nW, win_size, win_size, 1 attn_mask = input_mask_windows.view( -1, self.win_size * self.win_size ) # nW, win_size*win_size attn_mask = attn_mask.unsqueeze(2) * attn_mask.unsqueeze( 1 ) # nW, win_size*win_size, win_size*win_size attn_mask = attn_mask.masked_fill( attn_mask != 0, float(-100.0) ).masked_fill(attn_mask == 0, float(0.0)) else: attn_mask = None ## shift mask if self.shift_size > 0: # calculate attention mask for SW-MSA shift_mask = torch.zeros((1, H, W, 1)).type_as(x) h_slices = ( slice(0, -self.win_size), slice(-self.win_size, -self.shift_size), slice(-self.shift_size, None), ) w_slices = ( slice(0, -self.win_size), slice(-self.win_size, -self.shift_size), slice(-self.shift_size, None), ) cnt = 0 for h in h_slices: for w in w_slices: shift_mask[:, h, w, :] = cnt cnt += 1 shift_mask_windows = window_partition( shift_mask, self.win_size ) # nW, win_size, win_size, 1 shift_mask_windows = shift_mask_windows.view( -1, self.win_size * self.win_size ) # nW, win_size*win_size shift_attn_mask = shift_mask_windows.unsqueeze( 1 ) - shift_mask_windows.unsqueeze( 2 ) # nW, win_size*win_size, win_size*win_size shift_attn_mask = shift_attn_mask.masked_fill( shift_attn_mask != 0, float(-100.0) ).masked_fill(shift_attn_mask == 0, float(0.0)) attn_mask = ( attn_mask + shift_attn_mask if attn_mask is not None else shift_attn_mask ) attn_kv = attn_kv.view(B, H, W, C) # cyclic shift if self.shift_size > 0: shifted_kv = torch.roll( attn_kv, shifts=(-self.shift_size, -self.shift_size), dims=(1, 2) ) else: shifted_kv = attn_kv # partition windows attn_kv_windows = window_partition( shifted_kv, self.win_size ) # nW*B, win_size, win_size, C attn_kv_windows = attn_kv_windows.view( -1, self.win_size * self.win_size, C ) # nW*B, win_size*win_size, C x = x.view(B, H, W, C) # cyclic shift if self.shift_size > 0: shifted_x = torch.roll( x, shifts=(-self.shift_size, -self.shift_size), dims=(1, 2) ) else: shifted_x = x # partition windows x_windows = window_partition( shifted_x, self.win_size ) # nW*B, win_size, win_size, C x_windows = x_windows.view( -1, self.win_size * self.win_size, C ) # nW*B, win_size*win_size, C ### multi-head self-attention shortcut1 = x_windows # prenorm x_windows = self.norm1(x_windows) # W-MSA/SW-MSA attn_windows = self.attn( x_windows, mask=attn_mask ) # nW*B, win_size*win_size, C x_windows = shortcut1 + self.drop_path(attn_windows) ### multi-head cross-attention shortcut2 = x_windows # prenorm x_windows = self.norm2(x_windows) attn_kv_windows = self.norm_kv(attn_kv_windows) # W-MCA/SW-MCA attn_windows = self.cross_attn( x_windows, attn_kv=attn_kv_windows, mask=attn_mask ) # nW*B, win_size*win_size, C attn_windows = shortcut2 + self.drop_path(attn_windows) # merge windows attn_windows = attn_windows.view(-1, self.win_size, self.win_size, C) shifted_x = window_reverse(attn_windows, self.win_size, H, W) # B H' W' C # reverse cyclic shift if self.shift_size > 0: x = torch.roll( shifted_x, shifts=(self.shift_size, self.shift_size), dims=(1, 2) ) else: x = shifted_x x = x.view(B, H * W, C) # FFN x = x + self.drop_path(self.mlp(self.norm3(x))) del attn_mask return x def flops(self): flops = 0 H, W = self.input_resolution # norm1 flops += self.dim * H * W # W-MSA/SW-MSA flops += self.attn.flops(H, W) flops += self.cross_attn.flops(H, W) # norm2 flops += self.dim * H * W # mlp flops += self.mlp.flops(H, W) print("LeWin:{%.2f}" % (flops / 1e9)) return flops ########### LeWinTransformer_CatCross ############# class LeWinTransformer_CatCross(nn.Module): def __init__( self, dim, input_resolution, num_heads, win_size=8, shift_size=0, mlp_ratio=4.0, qkv_bias=True, qk_scale=None, drop=0.0, attn_drop=0.0, drop_path=0.0, act_layer=nn.GELU, norm_layer=nn.LayerNorm, token_projection="linear", token_mlp="ffn", ): super().__init__() self.dim = dim self.input_resolution = input_resolution self.num_heads = num_heads self.win_size = win_size self.shift_size = shift_size self.mlp_ratio = mlp_ratio if min(self.input_resolution) <= self.win_size: self.shift_size = 0 self.win_size = min(self.input_resolution) assert 0 <= self.shift_size < self.win_size, "shift_size must in 0-win_size" self.norm1 = norm_layer(dim) self.norm_kv = norm_layer(dim) self.cross_attn = WindowAttention( dim, win_size=to_2tuple(self.win_size), num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale, attn_drop=attn_drop, proj_drop=drop, token_projection="linear_concat", ) self.drop_path = DropPath(drop_path) if drop_path > 0.0 else nn.Identity() self.norm2 = norm_layer(dim) mlp_hidden_dim = int(dim * mlp_ratio) self.mlp = ( Mlp( in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop, ) if token_mlp == "ffn" else LeFF(dim, mlp_hidden_dim, act_layer=act_layer, drop=drop) ) def extra_repr(self) -> str: return ( f"dim={self.dim}, input_resolution={self.input_resolution}, num_heads={self.num_heads}, " f"win_size={self.win_size}, shift_size={self.shift_size}, mlp_ratio={self.mlp_ratio}" ) def forward(self, x, attn_kv=None, mask=None): B, L, C = x.shape H = int(math.sqrt(L)) W = int(math.sqrt(L)) ## input mask if mask != None: input_mask = F.interpolate(mask, size=(H, W)).permute(0, 2, 3, 1) input_mask_windows = window_partition( input_mask, self.win_size ) # nW, win_size, win_size, 1 attn_mask = input_mask_windows.view( -1, self.win_size * self.win_size ) # nW, win_size*win_size attn_mask = attn_mask.unsqueeze(2) * attn_mask.unsqueeze( 1 ) # nW, win_size*win_size, win_size*win_size attn_mask = attn_mask.masked_fill( attn_mask != 0, float(-100.0) ).masked_fill(attn_mask == 0, float(0.0)) else: attn_mask = None ## shift mask if self.shift_size > 0: # calculate attention mask for SW-MSA shift_mask = torch.zeros((1, H, W, 1)).type_as(x) h_slices = ( slice(0, -self.win_size), slice(-self.win_size, -self.shift_size), slice(-self.shift_size, None), ) w_slices = ( slice(0, -self.win_size), slice(-self.win_size, -self.shift_size), slice(-self.shift_size, None), ) cnt = 0 for h in h_slices: for w in w_slices: shift_mask[:, h, w, :] = cnt cnt += 1 shift_mask_windows = window_partition( shift_mask, self.win_size ) # nW, win_size, win_size, 1 shift_mask_windows = shift_mask_windows.view( -1, self.win_size * self.win_size ) # nW, win_size*win_size shift_attn_mask = shift_mask_windows.unsqueeze( 1 ) - shift_mask_windows.unsqueeze( 2 ) # nW, win_size*win_size, win_size*win_size shift_attn_mask = shift_attn_mask.masked_fill( shift_attn_mask != 0, float(-100.0) ).masked_fill(shift_attn_mask == 0, float(0.0)) attn_mask = ( attn_mask + shift_attn_mask if attn_mask is not None else shift_attn_mask ) attn_kv = attn_kv.view(B, H, W, C) # cyclic shift if self.shift_size > 0: shifted_kv = torch.roll( attn_kv, shifts=(-self.shift_size, -self.shift_size), dims=(1, 2) ) else: shifted_kv = attn_kv # partition windows attn_kv_windows = window_partition( shifted_kv, self.win_size ) # nW*B, win_size, win_size, C attn_kv_windows = attn_kv_windows.view( -1, self.win_size * self.win_size, C ) # nW*B, win_size*win_size, C x = x.view(B, H, W, C) # cyclic shift if self.shift_size > 0: shifted_x = torch.roll( x, shifts=(-self.shift_size, -self.shift_size), dims=(1, 2) ) else: shifted_x = x # partition windows x_windows = window_partition( shifted_x, self.win_size ) # nW*B, win_size, win_size, C x_windows = x_windows.view( -1, self.win_size * self.win_size, C ) # nW*B, win_size*win_size, C ### multi-head cross-attention shortcut1 = x_windows # prenorm x_windows = self.norm1(x_windows) attn_kv_windows = self.norm_kv(attn_kv_windows) # W-MCA/SW-MCA attn_windows = self.cross_attn( x_windows, attn_kv=attn_kv_windows, mask=attn_mask ) # nW*B, win_size*win_size, C attn_windows = shortcut1 + self.drop_path(attn_windows) # merge windows attn_windows = attn_windows.view(-1, self.win_size, self.win_size, C) shifted_x = window_reverse(attn_windows, self.win_size, H, W) # B H' W' C # reverse cyclic shift if self.shift_size > 0: x = torch.roll( shifted_x, shifts=(self.shift_size, self.shift_size), dims=(1, 2) ) else: x = shifted_x x = x.view(B, H * W, C) # FFN x = x + self.drop_path(self.mlp(self.norm2(x))) del attn_mask return x def flops(self): flops = 0 H, W = self.input_resolution # norm1 flops += self.dim * H * W # W-MSA/SW-MSA flops += self.cross_attn.flops(H, W) # norm2 flops += self.dim * H * W # mlp flops += self.mlp.flops(H, W) print("LeWin:{%.2f}" % (flops / 1e9)) return flops ######################################### ########### Basic layer of Uformer ################ class BasicUformerLayer(nn.Module): def __init__( self, dim, output_dim, input_resolution, depth, num_heads, win_size, mlp_ratio=4.0, qkv_bias=True, qk_scale=None, drop=0.0, attn_drop=0.0, drop_path=0.0, norm_layer=nn.LayerNorm, use_checkpoint=False, token_projection="linear", token_mlp="ffn", se_layer=False, ): super().__init__() self.dim = dim self.input_resolution = input_resolution self.depth = depth self.use_checkpoint = use_checkpoint # build blocks self.blocks = nn.ModuleList( [ LeWinTransformerBlock( dim=dim, input_resolution=input_resolution, num_heads=num_heads, win_size=win_size, shift_size=0 if (i % 2 == 0) else win_size // 2, mlp_ratio=mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop, attn_drop=attn_drop, drop_path=( drop_path[i] if isinstance(drop_path, list) else drop_path ), norm_layer=norm_layer, token_projection=token_projection, token_mlp=token_mlp, se_layer=se_layer, ) for i in range(depth) ] ) def extra_repr(self) -> str: return f"dim={self.dim}, input_resolution={self.input_resolution}, depth={self.depth}" def forward(self, x, mask=None): for blk in self.blocks: if self.use_checkpoint: x = checkpoint.checkpoint(blk, x) else: x = blk(x, mask) return x def flops(self): flops = 0 for blk in self.blocks: flops += blk.flops() return flops ########### Basic decoderlayer of Uformer_Cross ################ class CrossUformerLayer(nn.Module): def __init__( self, dim, output_dim, input_resolution, depth, num_heads, win_size, mlp_ratio=4.0, qkv_bias=True, qk_scale=None, drop=0.0, attn_drop=0.0, drop_path=0.0, norm_layer=nn.LayerNorm, use_checkpoint=False, token_projection="linear", token_mlp="ffn", ): super().__init__() self.dim = dim self.input_resolution = input_resolution self.depth = depth self.use_checkpoint = use_checkpoint # build blocks self.blocks = nn.ModuleList( [ LeWinTransformer_Cross( dim=dim, input_resolution=input_resolution, num_heads=num_heads, win_size=win_size, shift_size=0 if (i % 2 == 0) else win_size // 2, mlp_ratio=mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop, attn_drop=attn_drop, drop_path=( drop_path[i] if isinstance(drop_path, list) else drop_path ), norm_layer=norm_layer, token_projection=token_projection, token_mlp=token_mlp, ) for i in range(depth) ] ) def extra_repr(self) -> str: return f"dim={self.dim}, input_resolution={self.input_resolution}, depth={self.depth}" def forward(self, x, attn_kv=None, mask=None): for blk in self.blocks: if self.use_checkpoint: x = checkpoint.checkpoint(blk, x) else: x = blk(x, attn_kv, mask) return x def flops(self): flops = 0 for blk in self.blocks: flops += blk.flops() return flops ########### Basic decoderlayer of Uformer_CatCross ################ class CatCrossUformerLayer(nn.Module): def __init__( self, dim, output_dim, input_resolution, depth, num_heads, win_size, mlp_ratio=4.0, qkv_bias=True, qk_scale=None, drop=0.0, attn_drop=0.0, drop_path=0.0, norm_layer=nn.LayerNorm, use_checkpoint=False, token_projection="linear", token_mlp="ffn", ): super().__init__() self.dim = dim self.input_resolution = input_resolution self.depth = depth self.use_checkpoint = use_checkpoint # build blocks self.blocks = nn.ModuleList( [ LeWinTransformer_CatCross( dim=dim, input_resolution=input_resolution, num_heads=num_heads, win_size=win_size, shift_size=0 if (i % 2 == 0) else win_size // 2, mlp_ratio=mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop, attn_drop=attn_drop, drop_path=( drop_path[i] if isinstance(drop_path, list) else drop_path ), norm_layer=norm_layer, token_projection=token_projection, token_mlp=token_mlp, ) for i in range(depth) ] ) def extra_repr(self) -> str: return f"dim={self.dim}, input_resolution={self.input_resolution}, depth={self.depth}" def forward(self, x, attn_kv=None, mask=None): for blk in self.blocks: if self.use_checkpoint: x = checkpoint.checkpoint(blk, x) else: x = blk(x, attn_kv, mask) return x def flops(self): flops = 0 for blk in self.blocks: flops += blk.flops() return flops # @ARCH_REGISTRY.register() class Uformer(nn.Module): def __init__( self, img_size=128, img_ch=3, output_ch=3, embed_dim=32, depths=[2, 2, 2, 2, 2, 2, 2, 2, 2], num_heads=[1, 2, 4, 8, 16, 16, 8, 4, 2], win_size=8, mlp_ratio=4.0, qkv_bias=True, qk_scale=None, drop_rate=0.0, attn_drop_rate=0.0, drop_path_rate=0.1, norm_layer=nn.LayerNorm, patch_norm=True, use_checkpoint=False, token_projection="linear", token_mlp="ffn", se_layer=False, dowsample=Downsample, upsample=Upsample, **kwargs, ): super().__init__() in_chans = img_ch out_chans = output_ch self.num_enc_layers = len(depths) // 2 self.num_dec_layers = len(depths) // 2 self.embed_dim = embed_dim self.patch_norm = patch_norm self.mlp_ratio = mlp_ratio self.token_projection = token_projection self.mlp = token_mlp self.win_size = win_size self.reso = img_size self.pos_drop = nn.Dropout(p=drop_rate) # stochastic depth enc_dpr = [ x.item() for x in torch.linspace( 0, drop_path_rate, sum(depths[: self.num_enc_layers]) ) ] conv_dpr = [drop_path_rate] * depths[4] dec_dpr = enc_dpr[::-1] # build layers # Input/Output self.input_proj = InputProj( in_channel=in_chans, out_channel=embed_dim, kernel_size=3, stride=1, act_layer=nn.LeakyReLU, ) self.output_proj = OutputProj( in_channel=2 * embed_dim, out_channel=out_chans, kernel_size=3, stride=1 ) # Encoder self.encoderlayer_0 = BasicUformerLayer( dim=embed_dim, output_dim=embed_dim, input_resolution=(img_size, img_size), depth=depths[0], num_heads=num_heads[0], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=enc_dpr[sum(depths[:0]) : sum(depths[:1])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, se_layer=se_layer, ) self.dowsample_0 = dowsample(embed_dim, embed_dim * 2) self.encoderlayer_1 = BasicUformerLayer( dim=embed_dim * 2, output_dim=embed_dim * 2, input_resolution=(img_size // 2, img_size // 2), depth=depths[1], num_heads=num_heads[1], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=enc_dpr[sum(depths[:1]) : sum(depths[:2])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, se_layer=se_layer, ) self.dowsample_1 = dowsample(embed_dim * 2, embed_dim * 4) self.encoderlayer_2 = BasicUformerLayer( dim=embed_dim * 4, output_dim=embed_dim * 4, input_resolution=(img_size // (2**2), img_size // (2**2)), depth=depths[2], num_heads=num_heads[2], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=enc_dpr[sum(depths[:2]) : sum(depths[:3])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, se_layer=se_layer, ) self.dowsample_2 = dowsample(embed_dim * 4, embed_dim * 8) self.encoderlayer_3 = BasicUformerLayer( dim=embed_dim * 8, output_dim=embed_dim * 8, input_resolution=(img_size // (2**3), img_size // (2**3)), depth=depths[3], num_heads=num_heads[3], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=enc_dpr[sum(depths[:3]) : sum(depths[:4])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, se_layer=se_layer, ) self.dowsample_3 = dowsample(embed_dim * 8, embed_dim * 16) # Bottleneck self.conv = BasicUformerLayer( dim=embed_dim * 16, output_dim=embed_dim * 16, input_resolution=(img_size // (2**4), img_size // (2**4)), depth=depths[4], num_heads=num_heads[4], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=conv_dpr, norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, se_layer=se_layer, ) # Decoder self.upsample_0 = upsample(embed_dim * 16, embed_dim * 8) self.decoderlayer_0 = BasicUformerLayer( dim=embed_dim * 16, output_dim=embed_dim * 16, input_resolution=(img_size // (2**3), img_size // (2**3)), depth=depths[5], num_heads=num_heads[5], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dec_dpr[: depths[5]], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, se_layer=se_layer, ) self.upsample_1 = upsample(embed_dim * 16, embed_dim * 4) self.decoderlayer_1 = BasicUformerLayer( dim=embed_dim * 8, output_dim=embed_dim * 8, input_resolution=(img_size // (2**2), img_size // (2**2)), depth=depths[6], num_heads=num_heads[6], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dec_dpr[sum(depths[5:6]) : sum(depths[5:7])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, se_layer=se_layer, ) self.upsample_2 = upsample(embed_dim * 8, embed_dim * 2) self.decoderlayer_2 = BasicUformerLayer( dim=embed_dim * 4, output_dim=embed_dim * 4, input_resolution=(img_size // 2, img_size // 2), depth=depths[7], num_heads=num_heads[7], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dec_dpr[sum(depths[5:7]) : sum(depths[5:8])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, se_layer=se_layer, ) self.upsample_3 = upsample(embed_dim * 4, embed_dim) self.decoderlayer_3 = BasicUformerLayer( dim=embed_dim * 2, output_dim=embed_dim * 2, input_resolution=(img_size, img_size), depth=depths[8], num_heads=num_heads[8], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dec_dpr[sum(depths[5:8]) : sum(depths[5:9])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, se_layer=se_layer, ) self.activation = nn.Sequential(nn.Sigmoid()) self.apply(self._init_weights) def _init_weights(self, m): if isinstance(m, nn.Linear): trunc_normal_(m.weight, std=0.02) if isinstance(m, nn.Linear) and m.bias is not None: nn.init.constant_(m.bias, 0) elif isinstance(m, nn.LayerNorm): nn.init.constant_(m.bias, 0) nn.init.constant_(m.weight, 1.0) @torch.jit.ignore def no_weight_decay(self): return {"absolute_pos_embed"} @torch.jit.ignore def no_weight_decay_keywords(self): return {"relative_position_bias_table"} def extra_repr(self) -> str: return f"embed_dim={self.embed_dim}, token_projection={self.token_projection}, token_mlp={self.mlp},win_size={self.win_size}" def forward(self, x, mask=None): # Input Projection y = self.input_proj(x) y = self.pos_drop(y) # Encoder conv0 = self.encoderlayer_0(y, mask=mask) pool0 = self.dowsample_0(conv0) conv1 = self.encoderlayer_1(pool0, mask=mask) pool1 = self.dowsample_1(conv1) conv2 = self.encoderlayer_2(pool1, mask=mask) pool2 = self.dowsample_2(conv2) conv3 = self.encoderlayer_3(pool2, mask=mask) pool3 = self.dowsample_3(conv3) # Bottleneck conv4 = self.conv(pool3, mask=mask) # Decoder up0 = self.upsample_0(conv4) deconv0 = torch.cat([up0, conv3], -1) deconv0 = self.decoderlayer_0(deconv0, mask=mask) up1 = self.upsample_1(deconv0) deconv1 = torch.cat([up1, conv2], -1) deconv1 = self.decoderlayer_1(deconv1, mask=mask) up2 = self.upsample_2(deconv1) deconv2 = torch.cat([up2, conv1], -1) deconv2 = self.decoderlayer_2(deconv2, mask=mask) up3 = self.upsample_3(deconv2) deconv3 = torch.cat([up3, conv0], -1) deconv3 = self.decoderlayer_3(deconv3, mask=mask) # Output Projection y = self.output_proj(deconv3) y = self.activation(y) return y def flops(self): flops = 0 # Input Projection flops += self.input_proj.flops(self.reso, self.reso) # Encoder flops += self.encoderlayer_0.flops() + self.dowsample_0.flops( self.reso, self.reso ) flops += self.encoderlayer_1.flops() + self.dowsample_1.flops( self.reso // 2, self.reso // 2 ) flops += self.encoderlayer_2.flops() + self.dowsample_2.flops( self.reso // 2**2, self.reso // 2**2 ) flops += self.encoderlayer_3.flops() + self.dowsample_3.flops( self.reso // 2**3, self.reso // 2**3 ) # Bottleneck flops += self.conv.flops() # Decoder flops += ( self.upsample_0.flops(self.reso // 2**4, self.reso // 2**4) + self.decoderlayer_0.flops() ) flops += ( self.upsample_1.flops(self.reso // 2**3, self.reso // 2**3) + self.decoderlayer_1.flops() ) flops += ( self.upsample_2.flops(self.reso // 2**2, self.reso // 2**2) + self.decoderlayer_2.flops() ) flops += ( self.upsample_3.flops(self.reso // 2, self.reso // 2) + self.decoderlayer_3.flops() ) # Output Projection flops += self.output_proj.flops(self.reso, self.reso) return flops class Uformer_Cross(nn.Module): def __init__( self, img_size=128, in_chans=3, out_chans=3, embed_dim=32, depths=[2, 2, 2, 2, 2, 2, 2, 2, 2], num_heads=[1, 2, 4, 8, 16, 8, 4, 2, 1], win_size=8, mlp_ratio=4.0, qkv_bias=True, qk_scale=None, drop_rate=0.0, attn_drop_rate=0.0, drop_path_rate=0.1, norm_layer=nn.LayerNorm, patch_norm=True, use_checkpoint=False, token_projection="linear", token_mlp="ffn", dowsample=Downsample, upsample=Upsample, **kwargs, ): super().__init__() self.num_enc_layers = len(depths) // 2 self.num_dec_layers = len(depths) // 2 self.embed_dim = embed_dim self.patch_norm = patch_norm self.mlp_ratio = mlp_ratio self.token_projection = token_projection self.mlp = token_mlp self.win_size = win_size self.reso = img_size self.pos_drop = nn.Dropout(p=drop_rate) # stochastic depth enc_dpr = [ x.item() for x in torch.linspace( 0, drop_path_rate, sum(depths[: self.num_enc_layers]) ) ] conv_dpr = [drop_path_rate] * depths[4] dec_dpr = enc_dpr[::-1] # build layers # Input/Output self.input_proj = InputProj( in_channel=in_chans, out_channel=embed_dim, kernel_size=3, stride=1, act_layer=nn.LeakyReLU, ) self.output_proj = OutputProj( in_channel=embed_dim, out_channel=out_chans, kernel_size=3, stride=1 ) # Encoder self.encoderlayer_0 = BasicUformerLayer( dim=embed_dim, output_dim=embed_dim, input_resolution=(img_size, img_size), depth=depths[0], num_heads=num_heads[0], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=enc_dpr[sum(depths[:0]) : sum(depths[:1])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, ) self.dowsample_0 = dowsample(embed_dim, embed_dim * 2) self.encoderlayer_1 = BasicUformerLayer( dim=embed_dim * 2, output_dim=embed_dim * 2, input_resolution=(img_size // 2, img_size // 2), depth=depths[1], num_heads=num_heads[1], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=enc_dpr[sum(depths[:1]) : sum(depths[:2])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, ) self.dowsample_1 = dowsample(embed_dim * 2, embed_dim * 4) self.encoderlayer_2 = BasicUformerLayer( dim=embed_dim * 4, output_dim=embed_dim * 4, input_resolution=(img_size // (2**2), img_size // (2**2)), depth=depths[2], num_heads=num_heads[2], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=enc_dpr[sum(depths[:2]) : sum(depths[:3])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, ) self.dowsample_2 = dowsample(embed_dim * 4, embed_dim * 8) self.encoderlayer_3 = BasicUformerLayer( dim=embed_dim * 8, output_dim=embed_dim * 8, input_resolution=(img_size // (2**3), img_size // (2**3)), depth=depths[3], num_heads=num_heads[3], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=enc_dpr[sum(depths[:3]) : sum(depths[:4])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, ) self.dowsample_3 = dowsample(embed_dim * 8, embed_dim * 16) # Bottleneck self.conv = BasicUformerLayer( dim=embed_dim * 16, output_dim=embed_dim * 16, input_resolution=(img_size // (2**4), img_size // (2**4)), depth=depths[4], num_heads=num_heads[4], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=conv_dpr, norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, ) # Decoder self.upsample_0 = upsample(embed_dim * 16, embed_dim * 8) self.decoderlayer_0 = CrossUformerLayer( dim=embed_dim * 8, output_dim=embed_dim * 8, input_resolution=(img_size // (2**3), img_size // (2**3)), depth=depths[5], num_heads=num_heads[5], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dec_dpr[: depths[5]], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, ) self.upsample_1 = upsample(embed_dim * 8, embed_dim * 4) self.decoderlayer_1 = CrossUformerLayer( dim=embed_dim * 4, output_dim=embed_dim * 4, input_resolution=(img_size // (2**2), img_size // (2**2)), depth=depths[6], num_heads=num_heads[6], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dec_dpr[sum(depths[5:6]) : sum(depths[5:7])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, ) self.upsample_2 = upsample(embed_dim * 4, embed_dim * 2) self.decoderlayer_2 = CrossUformerLayer( dim=embed_dim * 2, output_dim=embed_dim * 2, input_resolution=(img_size // 2, img_size // 2), depth=depths[7], num_heads=num_heads[7], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dec_dpr[sum(depths[5:7]) : sum(depths[5:8])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, ) self.upsample_3 = upsample(embed_dim * 2, embed_dim) self.decoderlayer_3 = CrossUformerLayer( dim=embed_dim, output_dim=embed_dim, input_resolution=(img_size, img_size), depth=depths[8], num_heads=num_heads[8], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dec_dpr[sum(depths[5:8]) : sum(depths[5:9])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, ) self.apply(self._init_weights) def _init_weights(self, m): if isinstance(m, nn.Linear): trunc_normal_(m.weight, std=0.02) if isinstance(m, nn.Linear) and m.bias is not None: nn.init.constant_(m.bias, 0) elif isinstance(m, nn.LayerNorm): nn.init.constant_(m.bias, 0) nn.init.constant_(m.weight, 1.0) @torch.jit.ignore def no_weight_decay(self): return {"absolute_pos_embed"} @torch.jit.ignore def no_weight_decay_keywords(self): return {"relative_position_bias_table"} def extra_repr(self) -> str: return f"embed_dim={self.embed_dim}, token_projection={self.token_projection}, token_mlp={self.mlp},win_size={self.win_size}" def forward(self, x, mask=None): # Input Projection y = self.input_proj(x) y = self.pos_drop(y) # Encoder conv0 = self.encoderlayer_0(y, mask=mask) pool0 = self.dowsample_0(conv0) conv1 = self.encoderlayer_1(pool0, mask=mask) pool1 = self.dowsample_1(conv1) conv2 = self.encoderlayer_2(pool1, mask=mask) pool2 = self.dowsample_2(conv2) conv3 = self.encoderlayer_3(pool2, mask=mask) pool3 = self.dowsample_3(conv3) # Bottleneck conv4 = self.conv(pool3, mask=mask) # Decoder up0 = self.upsample_0(conv4) deconv0 = self.decoderlayer_0(up0, attn_kv=conv3, mask=mask) up1 = self.upsample_1(deconv0) deconv1 = self.decoderlayer_1(up1, attn_kv=conv2, mask=mask) up2 = self.upsample_2(deconv1) deconv2 = self.decoderlayer_2(up2, attn_kv=conv1, mask=mask) up3 = self.upsample_3(deconv2) deconv3 = self.decoderlayer_3(up3, attn_kv=conv0, mask=mask) # Output Projection y = self.output_proj(deconv3) return x + y def flops(self): flops = 0 # Input Projection flops += self.input_proj.flops(self.reso, self.reso) # Encoder flops += self.encoderlayer_0.flops() + self.dowsample_0.flops( self.reso, self.reso ) flops += self.encoderlayer_1.flops() + self.dowsample_1.flops( self.reso // 2, self.reso // 2 ) flops += self.encoderlayer_2.flops() + self.dowsample_2.flops( self.reso // 2**2, self.reso // 2**2 ) flops += self.encoderlayer_3.flops() + self.dowsample_3.flops( self.reso // 2**3, self.reso // 2**3 ) # Bottleneck flops += self.conv.flops() # Decoder flops += ( self.upsample_0.flops(self.reso // 2**4, self.reso // 2**4) + self.decoderlayer_0.flops() ) flops += ( self.upsample_1.flops(self.reso // 2**3, self.reso // 2**3) + self.decoderlayer_1.flops() ) flops += ( self.upsample_2.flops(self.reso // 2**2, self.reso // 2**2) + self.decoderlayer_2.flops() ) flops += ( self.upsample_3.flops(self.reso // 2, self.reso // 2) + self.decoderlayer_3.flops() ) # Output Projection flops += self.output_proj.flops(self.reso, self.reso) return flops class Uformer_CatCross(nn.Module): def __init__( self, img_size=128, in_chans=3, out_chans=3, embed_dim=32, depths=[2, 2, 2, 2, 2, 2, 2, 2, 2], num_heads=[1, 2, 4, 8, 16, 8, 4, 2, 1], win_size=8, mlp_ratio=4.0, qkv_bias=True, qk_scale=None, drop_rate=0.0, attn_drop_rate=0.0, drop_path_rate=0.1, norm_layer=nn.LayerNorm, patch_norm=True, use_checkpoint=False, token_projection="linear", token_mlp="ffn", dowsample=Downsample, upsample=Upsample, **kwargs, ): super().__init__() self.num_enc_layers = len(depths) // 2 self.num_dec_layers = len(depths) // 2 self.embed_dim = embed_dim self.patch_norm = patch_norm self.mlp_ratio = mlp_ratio self.token_projection = token_projection self.mlp = token_mlp self.win_size = win_size self.reso = img_size self.pos_drop = nn.Dropout(p=drop_rate) # stochastic depth enc_dpr = [ x.item() for x in torch.linspace( 0, drop_path_rate, sum(depths[: self.num_enc_layers]) ) ] # stochastic depth decay rule conv_dpr = [drop_path_rate] * depths[4] dec_dpr = enc_dpr[::-1] # build layers # Input/Output self.input_proj = InputProj( in_channel=in_chans, out_channel=embed_dim, kernel_size=3, stride=1, act_layer=nn.LeakyReLU, ) self.output_proj = OutputProj( in_channel=embed_dim, out_channel=out_chans, kernel_size=3, stride=1 ) # Encoder self.encoderlayer_0 = BasicUformerLayer( dim=embed_dim, output_dim=embed_dim, input_resolution=(img_size, img_size), depth=depths[0], num_heads=num_heads[0], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=enc_dpr[sum(depths[:0]) : sum(depths[:1])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, ) self.dowsample_0 = dowsample(embed_dim, embed_dim * 2) self.encoderlayer_1 = BasicUformerLayer( dim=embed_dim * 2, output_dim=embed_dim * 2, input_resolution=(img_size // 2, img_size // 2), depth=depths[1], num_heads=num_heads[1], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=enc_dpr[sum(depths[:1]) : sum(depths[:2])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, ) self.dowsample_1 = dowsample(embed_dim * 2, embed_dim * 4) self.encoderlayer_2 = BasicUformerLayer( dim=embed_dim * 4, output_dim=embed_dim * 4, input_resolution=(img_size // (2**2), img_size // (2**2)), depth=depths[2], num_heads=num_heads[2], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=enc_dpr[sum(depths[:2]) : sum(depths[:3])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, ) self.dowsample_2 = dowsample(embed_dim * 4, embed_dim * 8) self.encoderlayer_3 = BasicUformerLayer( dim=embed_dim * 8, output_dim=embed_dim * 8, input_resolution=(img_size // (2**3), img_size // (2**3)), depth=depths[3], num_heads=num_heads[3], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=enc_dpr[sum(depths[:3]) : sum(depths[:4])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, ) self.dowsample_3 = dowsample(embed_dim * 8, embed_dim * 16) # Bottleneck self.conv = BasicUformerLayer( dim=embed_dim * 16, output_dim=embed_dim * 16, input_resolution=(img_size // (2**4), img_size // (2**4)), depth=depths[4], num_heads=num_heads[4], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=conv_dpr, norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, ) # Decoder self.upsample_0 = upsample(embed_dim * 16, embed_dim * 8) self.decoderlayer_0 = CatCrossUformerLayer( dim=embed_dim * 8, output_dim=embed_dim * 8, input_resolution=(img_size // (2**3), img_size // (2**3)), depth=depths[5], num_heads=num_heads[5], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dec_dpr[: depths[5]], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, ) self.upsample_1 = upsample(embed_dim * 8, embed_dim * 4) self.decoderlayer_1 = CatCrossUformerLayer( dim=embed_dim * 4, output_dim=embed_dim * 4, input_resolution=(img_size // (2**2), img_size // (2**2)), depth=depths[6], num_heads=num_heads[6], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dec_dpr[sum(depths[5:6]) : sum(depths[5:7])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, ) self.upsample_2 = upsample(embed_dim * 4, embed_dim * 2) self.decoderlayer_2 = CatCrossUformerLayer( dim=embed_dim * 2, output_dim=embed_dim * 2, input_resolution=(img_size // 2, img_size // 2), depth=depths[7], num_heads=num_heads[7], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dec_dpr[sum(depths[5:7]) : sum(depths[5:8])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, ) self.upsample_3 = upsample(embed_dim * 2, embed_dim) self.decoderlayer_3 = CatCrossUformerLayer( dim=embed_dim, output_dim=embed_dim, input_resolution=(img_size, img_size), depth=depths[8], num_heads=num_heads[8], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dec_dpr[sum(depths[5:8]) : sum(depths[5:9])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, ) self.apply(self._init_weights) def _init_weights(self, m): if isinstance(m, nn.Linear): trunc_normal_(m.weight, std=0.02) if isinstance(m, nn.Linear) and m.bias is not None: nn.init.constant_(m.bias, 0) elif isinstance(m, nn.LayerNorm): nn.init.constant_(m.bias, 0) nn.init.constant_(m.weight, 1.0) @torch.jit.ignore def no_weight_decay(self): return {"absolute_pos_embed"} @torch.jit.ignore def no_weight_decay_keywords(self): return {"relative_position_bias_table"} def extra_repr(self) -> str: return f"embed_dim={self.embed_dim}, token_projection={self.token_projection}, token_mlp={self.mlp},win_size={self.win_size}" def forward(self, x, mask=None): # Input Projection y = self.input_proj(x) y = self.pos_drop(y) # Encoder conv0 = self.encoderlayer_0(y, mask=mask) pool0 = self.dowsample_0(conv0) conv1 = self.encoderlayer_1(pool0, mask=mask) pool1 = self.dowsample_1(conv1) conv2 = self.encoderlayer_2(pool1, mask=mask) pool2 = self.dowsample_2(conv2) conv3 = self.encoderlayer_3(pool2, mask=mask) pool3 = self.dowsample_3(conv3) # Bottleneck conv4 = self.conv(pool3, mask=mask) # Decoder up0 = self.upsample_0(conv4) deconv0 = self.decoderlayer_0(up0, attn_kv=conv3, mask=mask) up1 = self.upsample_1(deconv0) deconv1 = self.decoderlayer_1(up1, attn_kv=conv2, mask=mask) up2 = self.upsample_2(deconv1) deconv2 = self.decoderlayer_2(up2, attn_kv=conv1, mask=mask) up3 = self.upsample_3(deconv2) deconv3 = self.decoderlayer_3(up3, attn_kv=conv0, mask=mask) # Output Projection y = self.output_proj(deconv3) return x + y def flops(self): flops = 0 # Input Projection flops += self.input_proj.flops(self.reso, self.reso) # Encoder flops += self.encoderlayer_0.flops() + self.dowsample_0.flops( self.reso, self.reso ) flops += self.encoderlayer_1.flops() + self.dowsample_1.flops( self.reso // 2, self.reso // 2 ) flops += self.encoderlayer_2.flops() + self.dowsample_2.flops( self.reso // 2**2, self.reso // 2**2 ) flops += self.encoderlayer_3.flops() + self.dowsample_3.flops( self.reso // 2**3, self.reso // 2**3 ) # Bottleneck flops += self.conv.flops() # Decoder flops += ( self.upsample_0.flops(self.reso // 2**4, self.reso // 2**4) + self.decoderlayer_0.flops() ) flops += ( self.upsample_1.flops(self.reso // 2**3, self.reso // 2**3) + self.decoderlayer_1.flops() ) flops += ( self.upsample_2.flops(self.reso // 2**2, self.reso // 2**2) + self.decoderlayer_2.flops() ) flops += ( self.upsample_3.flops(self.reso // 2, self.reso // 2) + self.decoderlayer_3.flops() ) # Output Projection flops += self.output_proj.flops(self.reso, self.reso) return flops # class LeWinformer(nn.Module): # def __init__(self, img_size=128, in_chans=3, # embed_dim=32, depth=12, # win_size=8, mlp_ratio=4., qkv_bias=True, qk_scale=None, # drop_rate=0., attn_drop_rate=0., drop_path_rate=0.1, # norm_layer=nn.LayerNorm, patch_norm=True, # use_checkpoint=False, token_projection='linear', token_mlp='ffn', se_layer=False,**kwargs): # super().__init__() # self.transformer_layers = nn.ModuleList([]) # self.embed_dim = embed_dim # self.num_heads = embed_dim//32 or 1 # self.patch_norm = patch_norm # self.mlp_ratio = mlp_ratio # self.token_projection = token_projection # self.mlp = token_mlp # self.win_size =win_size # self.pos_drop = nn.Dropout(p=drop_rate) # # stochastic depth # dpr = [x.item() for x in torch.linspace(0, drop_path_rate, depth)] # # build layers # # Input/Output # self.input_proj = InputProj(in_channel=in_chans, out_channel=embed_dim, kernel_size=3, stride=1, act_layer=nn.LeakyReLU) # self.output_proj = OutputProj(in_channel=embed_dim, out_channel=in_chans, kernel_size=3, stride=1) # # LeWin Transformer # for i in range(depth): # dim = embed_dim # self.transformer_layers.append(nn.ModuleList([BasicUformerLayer(dim=dim, # output_dim=embed_dim, # input_resolution=(img_size, # img_size), # depth=1, # num_heads=self.num_heads, # win_size=win_size, # mlp_ratio=self.mlp_ratio, # qkv_bias=qkv_bias, qk_scale=qk_scale, # drop=drop_rate, attn_drop=attn_drop_rate, # drop_path=dpr[i], # norm_layer=norm_layer, # use_checkpoint=use_checkpoint, # token_projection=token_projection,token_mlp=token_mlp,se_layer=se_layer), # Downsample() # ])) # self.apply(self._init_weights) # def _init_weights(self, m): # if isinstance(m, nn.Linear): # trunc_normal_(m.weight, std=.02) # if isinstance(m, nn.Linear) and m.bias is not None: # nn.init.constant_(m.bias, 0) # elif isinstance(m, nn.LayerNorm): # nn.init.constant_(m.bias, 0) # nn.init.constant_(m.weight, 1.0) # @torch.jit.ignore # def no_weight_decay(self): # return {'absolute_pos_embed'} # @torch.jit.ignore # def no_weight_decay_keywords(self): # return {'relative_position_bias_table'} # def extra_repr(self) -> str: # return f"embed_dim={self.embed_dim}, token_projection={self.token_projection}, token_mlp={self.mlp},win_size={self.win_size}" # def forward(self, x, mask=None): # # Input Projection # y = self.input_proj(x) # y = self.pos_drop(y) # #Encoder # for lewin in self.transformer_layers: # y = lewin(y) # # Output Projection # y = self.output_proj(y) # return x + y ### single-scale Uformer is computationally too costly. class Uformer_singlescale(nn.Module): def __init__( self, img_size=128, in_chans=3, out_chans=3, embed_dim=32, depths=[2, 2, 2, 2, 2, 2, 2, 2, 2], num_heads=[1, 2, 4, 8, 16, 16, 8, 4, 2], win_size=8, mlp_ratio=4.0, qkv_bias=True, qk_scale=None, drop_rate=0.0, attn_drop_rate=0.0, drop_path_rate=0.1, norm_layer=nn.LayerNorm, patch_norm=True, use_checkpoint=False, token_projection="linear", token_mlp="ffn", se_layer=False, downsample=Downsample, upsample=Upsample, **kwargs, ): super().__init__() self.num_enc_layers = len(depths) // 2 self.num_dec_layers = len(depths) // 2 self.embed_dim = embed_dim self.patch_norm = patch_norm self.mlp_ratio = mlp_ratio self.token_projection = token_projection self.mlp = token_mlp self.win_size = win_size self.pos_drop = nn.Dropout(p=drop_rate) # stochastic depth enc_dpr = [ x.item() for x in torch.linspace( 0, drop_path_rate, sum(depths[: self.num_enc_layers]) ) ] conv_dpr = [drop_path_rate] * depths[4] dec_dpr = enc_dpr[::-1] # build layers # Input/Output self.input_proj = InputProj( in_channel=in_chans, out_channel=embed_dim, kernel_size=3, stride=1, act_layer=nn.LeakyReLU, ) self.output_proj = OutputProj( in_channel=2 * embed_dim, out_channel=out_chans, kernel_size=3, stride=1 ) # Encoder self.encoderlayer_0 = BasicUformerLayer( dim=embed_dim, output_dim=embed_dim, input_resolution=(img_size, img_size), depth=depths[0], num_heads=num_heads[0], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=enc_dpr[sum(depths[:0]) : sum(depths[:1])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, se_layer=se_layer, ) self.downsample_0 = downsample(embed_dim, embed_dim * 2, downsample=False) self.encoderlayer_1 = BasicUformerLayer( dim=embed_dim * 2, output_dim=embed_dim * 2, input_resolution=(img_size, img_size), depth=depths[1], num_heads=num_heads[1], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=enc_dpr[sum(depths[:1]) : sum(depths[:2])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, se_layer=se_layer, ) self.downsample_1 = downsample(embed_dim * 2, embed_dim * 4, downsample=False) self.encoderlayer_2 = BasicUformerLayer( dim=embed_dim * 4, output_dim=embed_dim * 4, input_resolution=(img_size, img_size), depth=depths[2], num_heads=num_heads[2], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=enc_dpr[sum(depths[:2]) : sum(depths[:3])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, se_layer=se_layer, ) self.downsample_2 = downsample(embed_dim * 4, embed_dim * 8, downsample=False) self.encoderlayer_3 = BasicUformerLayer( dim=embed_dim * 8, output_dim=embed_dim * 8, input_resolution=(img_size, img_size), depth=depths[3], num_heads=num_heads[3], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=enc_dpr[sum(depths[:3]) : sum(depths[:4])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, se_layer=se_layer, ) self.downsample_3 = downsample(embed_dim * 8, embed_dim * 16, downsample=False) # Bottleneck self.conv = BasicUformerLayer( dim=embed_dim * 16, output_dim=embed_dim * 16, input_resolution=(img_size, img_size), depth=depths[4], num_heads=num_heads[4], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=conv_dpr, norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, se_layer=se_layer, ) # Decoder self.upsample_0 = upsample(embed_dim * 16, embed_dim * 8, upsample=False) self.decoderlayer_0 = BasicUformerLayer( dim=embed_dim * 16, output_dim=embed_dim * 16, input_resolution=(img_size, img_size), depth=depths[5], num_heads=num_heads[5], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dec_dpr[: depths[5]], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, se_layer=se_layer, ) self.upsample_1 = upsample(embed_dim * 16, embed_dim * 4, upsample=False) self.decoderlayer_1 = BasicUformerLayer( dim=embed_dim * 8, output_dim=embed_dim * 8, input_resolution=(img_size, img_size), depth=depths[6], num_heads=num_heads[6], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dec_dpr[sum(depths[5:6]) : sum(depths[5:7])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, se_layer=se_layer, ) self.upsample_2 = upsample(embed_dim * 8, embed_dim * 2, upsample=False) self.decoderlayer_2 = BasicUformerLayer( dim=embed_dim * 4, output_dim=embed_dim * 4, input_resolution=(img_size, img_size), depth=depths[7], num_heads=num_heads[7], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dec_dpr[sum(depths[5:7]) : sum(depths[5:8])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, se_layer=se_layer, ) self.upsample_3 = upsample(embed_dim * 4, embed_dim, upsample=False) self.decoderlayer_3 = BasicUformerLayer( dim=embed_dim * 2, output_dim=embed_dim * 2, input_resolution=(img_size, img_size), depth=depths[8], num_heads=num_heads[8], win_size=win_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dec_dpr[sum(depths[5:8]) : sum(depths[5:9])], norm_layer=norm_layer, use_checkpoint=use_checkpoint, token_projection=token_projection, token_mlp=token_mlp, se_layer=se_layer, ) self.apply(self._init_weights) def _init_weights(self, m): if isinstance(m, nn.Linear): trunc_normal_(m.weight, std=0.02) if isinstance(m, nn.Linear) and m.bias is not None: nn.init.constant_(m.bias, 0) elif isinstance(m, nn.LayerNorm): nn.init.constant_(m.bias, 0) nn.init.constant_(m.weight, 1.0) @torch.jit.ignore def no_weight_decay(self): return {"absolute_pos_embed"} @torch.jit.ignore def no_weight_decay_keywords(self): return {"relative_position_bias_table"} def extra_repr(self) -> str: return f"embed_dim={self.embed_dim}, token_projection={self.token_projection}, token_mlp={self.mlp},win_size={self.win_size}" def forward(self, x, mask=None): # Input Projection y = self.input_proj(x) y = self.pos_drop(y) # Encoder conv0 = self.encoderlayer_0(y, mask=mask) pool0 = self.downsample_0(conv0) conv1 = self.encoderlayer_1(pool0, mask=mask) pool1 = self.downsample_1(conv1) conv2 = self.encoderlayer_2(pool1, mask=mask) pool2 = self.downsample_2(conv2) conv3 = self.encoderlayer_3(pool2, mask=mask) pool3 = self.downsample_3(conv3) # Bottleneck conv4 = self.conv(pool3, mask=mask) # Decoder up0 = self.upsample_0(conv4) deconv0 = torch.cat([up0, conv3], -1) deconv0 = self.decoderlayer_0(deconv0, mask=mask) up1 = self.upsample_1(deconv0) deconv1 = torch.cat([up1, conv2], -1) deconv1 = self.decoderlayer_1(deconv1, mask=mask) up2 = self.upsample_2(deconv1) deconv2 = torch.cat([up2, conv1], -1) deconv2 = self.decoderlayer_2(deconv2, mask=mask) up3 = self.upsample_3(deconv2) deconv3 = torch.cat([up3, conv0], -1) deconv3 = self.decoderlayer_3(deconv3, mask=mask) # Output Projection y = self.output_proj(deconv3) return x + y if __name__ == "__main__": arch = Uformer input_size = 256 # arch = Uformer_Cross depths = [2, 2, 2, 2, 2, 2, 2, 2, 2] # model_restoration = UNet(dim=32) model_restoration = arch( img_size=input_size, embed_dim=44, depths=depths, win_size=8, mlp_ratio=4.0, qkv_bias=True, token_projection="linear", token_mlp="leff", downsample=Downsample, upsample=Upsample, se_layer=False, ) # arch = LeWinformer # depth = 20 # model_restoration = arch(embed_dim=16,depth=depth, # win_size=8, mlp_ratio=4., qkv_bias=True, # token_projection='linear', token_mlp='leff',se_layer=False) # from ptflops import get_model_complexity_info # macs, params = get_model_complexity_info(model_restoration, (3, input_size, input_size), as_strings=True, # print_per_layer_stat=True, verbose=True) # print('{:<30} {:<8}'.format('Computational complexity: ', macs)) # print('{:<30} {:<8}'.format('Number of parameters: ', params)) # print("number of GFLOPs: %.2f G"%(model_restoration.flops(input_size,input_size) / 1e9)) print("number of GFLOPs: %.2f G" % (model_restoration.flops() / 1e9))