Swin-Transformer通用视觉骨干网络主体结构代码解释
2024-05-11 14:17:49
Swin-Transformer由MSRA视觉计算组的team于2021年发表的工作，在多个视觉任务以及多个数据集上均取得了十分优秀的结果。在这里，我贴出我对于Swin-Transformer主体结构的一些代码的解释和tensor的shape的改变，由于时间的原因，可能会出现许多纰漏，希望大家多多指教
paper：https://arxiv.org/pdf/2111.09883v1.pdf
code：GitHub - microsoft/Swin-Transformer: This is an official implementation for "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows".
import torch
import torch.nn as nn
import torch.utils.checkpoint as checkpoint
from timm.models.layers import DropPath, to_2tuple, trunc_normal_
#-------------------------------#
# 此为对于MLP模块的定义
#-------------------------------#
class Mlp(nn.Module):def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.):super(Mlp,self).__init__()#---------------------------------------## 在这里使用Dropout的作用在于# 降低因为Linear层的使用所造成的过拟合现象的发生#---------------------------------------#out_features = out_features or in_featureshidden_features = hidden_features or in_featuresself.fc1 = nn.Linear(in_features, hidden_features)self.act = act_layer()self.fc2 = nn.Linear(hidden_features, out_features)self.drop = nn.Dropout(drop)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#--------------------------------------------------#
# 对于window_partition的定义
# 在这里x是一个tensor
# window size由自己定义
# view()的作用相当于numpy中的reshape,重新定义矩阵的形状
#--------------------------------------------------#
def window_partition(x, window_size):"""Args:x: (B, H, W, C)window_size (int): window sizeReturns:windows: (num_windows**2*B, window_size, window_size, C)"""B, H, W, C = x.shapex = x.view(B, H // window_size, window_size, W // window_size, window_size, C)#-------------------------------------------------------------------------------------------## 1. shape = B, H // window_size, W // window_size, window_size, window_size, C# 2. shape = num_windows**2*B, window_size, window_size, C(返回值window的shape)# 3. view(-1, window_size, window_size, C)的含义为 后三维度已经确定,第一维由整体矩阵根据后三个维度得到# 4. H // window_size = W // window_size即为 num_windows#-------------------------------------------------------------------------------------------#windows = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(-1, window_size, window_size, C)return windows#-----------------------------------#
# 此为对于window_reverse函数的定义
#-----------------------------------#
def window_reverse(windows, window_size, H, W):"""Args:windows: (num_windows**2*B, window_size, window_size, C)window_size (int): Window sizeH (int): Height of imageW (int): Width of imageReturns:x: (B, H, W, C)"""B = int(windows.shape[0] / (H * W / window_size / window_size))x = windows.view(B, H // window_size, W // window_size, window_size, window_size, -1)x = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(B, H, W, -1)return x#--------------------------------#
# 对于WindowAttention这个类的定义
# 该类支持滑动的以及未滑动的窗口图像
#--------------------------------#
class WindowAttention(nn.Module):r""" Window based multi-head self attention (W-MSA) module with relative position bias.It supports both of shifted and non-shifted window.Args:dim (int): Number of input channels.window_size (tuple[int]): The height and width of the window.num_heads (int): Number of attention heads.qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: Trueqk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if setattn_drop (float, optional): Dropout ratio of attention weight. Default: 0.0proj_drop (float, optional): Dropout ratio of output. Default: 0.0"""def __init__(self, dim, window_size, num_heads, qkv_bias=True, qk_scale=None, attn_drop=0., proj_drop=0.):super(WindowAttention,self).__init__()head_dim = dim // num_headsself.dim = dimself.window_size = window_size # Wh,Wwself.num_heads = num_headsself.scale = qk_scale or head_dim ** -0.5#----------------------------------------------------------------## define a parameter table of relative position bias# return: 2*Wh-1 * 2*Ww-1, nH# nn.Parameter作用为定义这些参数是可以学习的参数# torch.zeros()：其形状由变量参数size定义,返回一个由标量值0填充的张量#----------------------------------------------------------------#self.relative_position_bias_table = nn.Parameter(torch.zeros((2 * window_size[0] - 1) * (2 * window_size[1] - 1), num_heads))#----------------------------------------------------------------------------## get pair-wise relative position index for each token inside the window# torch.arange()所创建的张量是int类型# torch.meshgrid()的作用在于将两个类型相同的张量生成一个tensor矩阵# 这个矩阵的行数为第一个input_tensor的维度,列数为第二个input_tensor的维度# 之后又进行了stack操作,又增加了一个维度,所以此时的shape为2, Wh, Ww# 之后通过flatten将后两个维度压缩成一个维度，此时的shape为2, Wh*Ww#----------------------------------------------------------------------------#coords_h = torch.arange(self.window_size[0])coords_w = torch.arange(self.window_size[1])coords = torch.stack(torch.meshgrid([coords_h, coords_w]))coords_flatten = torch.flatten(coords, 1)#----------------------------------------------## [:, :, None]其中的None代表增加一个维度,具体的值为1# relative_coords的shape为2, Wh*Ww, Wh*Ww# 之后对其进行转置,此时的shape为Wh*Ww, Wh*Ww, 2#----------------------------------------------#relative_coords = coords_flatten[:, :, None] - coords_flatten[:, None, :]relative_coords = relative_coords.permute(1, 2, 0).contiguous()relative_coords[:, :, 0] += self.window_size[0] - 1 # shift to start from 0relative_coords[:, :, 1] += self.window_size[1] - 1relative_coords[:, :, 0] *= 2 * self.window_size[1] - 1#---------------------------------------------## relative_position_index的shape为Wh*Ww, Wh*Ww# 之后将它作为一个模型的常数#----------------------------------------- ---#relative_position_index = relative_coords.sum(-1)self.register_buffer("relative_position_index", relative_position_index)#--------------------------------------## 定义qkv以及proj,在他们之后均有一个Dropout# 以降低过拟合发生的风险#--------------------------------------#self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)self.attn_drop = nn.Dropout(attn_drop)self.proj = nn.Linear(dim, dim)self.proj_drop = nn.Dropout(proj_drop)#---------------------------------## 利用正太分布来生成一个点# 之后又定义了一个softmax分类器#---------------------------------#trunc_normal_(self.relative_position_bias_table, std=.02)self.softmax = nn.Softmax(dim=-1)#------------------------------------------------------## 该类前向传播函数的定义# input-x即为shape：num_windows**2*B,N(windows_size), C#------------------------------------------------------#def forward(self, x, mask=None):"""Args:x: input features with shape of (num_windows*B, N, C)mask: (0/-inf) mask with shape of (num_windows, Wh*Ww, Wh*Ww) or None"""B_, N, C = x.shape#--------------------------------------------------## 首先给qkv添加几个维度,之后再进行转置# shape= 3,B_,self.num_heads,N,C // self.num_heads# 之后便可以得知：q=3, k=B_, v=self.num_heads# 之后再求解q的值,最后进行注意力的计算#--------------------------------------------------#qkv = self.qkv(x).reshape(B_, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)q, k, v = qkv[0], qkv[1], qkv[2]q = q * self.scaleattn = (q @ k.transpose(-2, -1))#-----------------------------------------------## relative_position_bias的shape为Wh*Ww,Wh*Ww,nH# 之后又进行转置,此时的shape为nH, Wh*Ww, Wh*Ww# 之后又再第一维增加了一个维度#-----------------------------------------------#relative_position_bias = self.relative_position_bias_table[self.relative_position_index.view(-1)].view(self.window_size[0] * self.window_size[1],self.window_size[0] * self.window_size[1], -1)relative_position_bias = relative_position_bias.permute(2, 0, 1).contiguous()attn = attn + relative_position_bias.unsqueeze(0)#----------------------------## 在这里我们的mask的值为None# 所以直接pass through softmax# 之后过dropout#----------------------------#if mask is not None:nW = mask.shape[0]attn = attn.view(B_ // nW, nW, self.num_heads, N, N) + mask.unsqueeze(1).unsqueeze(0)attn = attn.view(-1, self.num_heads, N, N)attn = self.softmax(attn)else:attn = self.softmax(attn)attn = self.attn_drop(attn)#------------------------## 进行注意力加权运算# 之后过投影层# 最后过投影层的dropout#------------------------#x = (attn @ v).transpose(1, 2).reshape(B_, N, C)x = self.proj(x)x = self.proj_drop(x)return x#------------------------------## extra_repr以及flops函数的定义#------------------------------#def extra_repr(self) -> str:return f'dim={self.dim}, window_size={self.window_size}, num_heads={self.num_heads}'def flops(self, N):# calculate flops for 1 window with token length of Nflops = 0# qkv = self.qkv(x)flops += N * self.dim * 3 * self.dim# attn = (q @ k.transpose(-2, -1))flops += self.num_heads * N * (self.dim // self.num_heads) * N# x = (attn @ v)flops += self.num_heads * N * N * (self.dim // self.num_heads)# x = self.proj(x)flops += N * self.dim * self.dimreturn flopsclass SwinTransformerBlock(nn.Module):r""" Swin Transformer Block.Args:dim (int): Number of input channels. 输入图像的通道数input_resolution (tuple[int]): Input resulotion. 输入图像的分辨率num_heads (int): Number of attention heads. 多头注意力中注意力头的数目window_size (int): Window size.shift_size (int): Shift size for SW-MSA. 一次滑动操作所滑动的尺寸mlp_ratio (float): Ratio of mlp hidden dim to embedding dim.qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: Trueqk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set.drop (float, optional): Dropout rate. Default: 0.0attn_drop (float, optional): Attention dropout rate. Default: 0.0drop_path (float, optional): Stochastic depth rate. Default: 0.0act_layer (nn.Module, optional): Activation layer. Default: nn.GELUnorm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm"""def __init__(self, dim, input_resolution, num_heads, window_size=7, shift_size=0,mlp_ratio=4., qkv_bias=True, qk_scale=None, drop=0., attn_drop=0., drop_path=0.,act_layer=nn.GELU, norm_layer=nn.LayerNorm):super(SwinTransformerBlock,self).__init__()self.dim = dimself.input_resolution = input_resolutionself.num_heads = num_headsself.window_size = window_sizeself.shift_size = shift_sizeself.mlp_ratio = mlp_ratio#--------------------------------## 若input的尺寸小于窗口的大小#--------------------------------#if min(self.input_resolution) <= self.window_size:# if window size is larger than input resolution, we don't partition windowsself.shift_size = 0self.window_size = min(self.input_resolution)assert 0 <= self.shift_size < self.window_size, "shift_size must in 0-window_size"#-------------------------------------## 定义使用LayerNorm并且定义了窗口注意力机制# nn.Identity()相当于pass#-------------------------------------#self.norm1 = norm_layer(dim)self.attn = WindowAttention(dim, window_size=to_2tuple(self.window_size), num_heads=num_heads,qkv_bias=qkv_bias, qk_scale=qk_scale, attn_drop=attn_drop, proj_drop=drop)self.drop_path = DropPath(drop_path) if drop_path > 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 self.shift_size > 0:# calculate attention mask for SW-MSA# slice用于对数组元素进行截取，返回值为截取元素组成的一个新数组H, W = self.input_resolutionimg_mask = torch.zeros((1, H, W, 1)) # 1 H W 1h_slices = (slice(0, -self.window_size),slice(-self.window_size, -self.shift_size),slice(-self.shift_size, None))w_slices = (slice(0, -self.window_size),slice(-self.window_size, -self.shift_size),slice(-self.shift_size, None))cnt = 0#---------------------------------------------## 根据h_slices以及w_slices求取cnt的值并进行相关操作#---------------------------------------------#for h in h_slices:for w in w_slices:img_mask[:, h, w, :] = cntcnt += 1mask_windows = window_partition(img_mask, self.window_size) # nW, window_size, window_size, 1mask_windows = mask_windows.view(-1, self.window_size * self.window_size)attn_mask = mask_windows.unsqueeze(1) - mask_windows.unsqueeze(2)attn_mask = attn_mask.masked_fill(attn_mask != 0, float(-100.0)).masked_fill(attn_mask == 0, float(0.0))else:attn_mask = Noneself.register_buffer("attn_mask", attn_mask)def forward(self, x):H, W = self.input_resolutionB, L, C = x.shapeassert L == H * W, "input feature has wrong size"shortcut = xx = self.norm1(x)x = x.view(B, H, W, C)# cyclic shiftif self.shift_size > 0:shifted_x = torch.roll(x, shifts=(-self.shift_size, -self.shift_size), dims=(1, 2))else:shifted_x = x# partition windowsx_windows = window_partition(shifted_x, self.window_size) # nW*B, window_size, window_size, Cx_windows = x_windows.view(-1, self.window_size * self.window_size, C) # nW*B, window_size*window_size, C# W-MSA/SW-MSAattn_windows = self.attn(x_windows, mask=self.attn_mask) # nW*B, window_size*window_size, C# merge windowsattn_windows = attn_windows.view(-1, self.window_size, self.window_size, C)shifted_x = window_reverse(attn_windows, self.window_size, H, W) # B H' W' C# reverse cyclic shiftif self.shift_size > 0:x = torch.roll(shifted_x, shifts=(self.shift_size, self.shift_size), dims=(1, 2))else:x = shifted_xx = x.view(B, H * W, C)# FFNx = shortcut + self.drop_path(x)x = x + self.drop_path(self.mlp(self.norm2(x)))return xdef extra_repr(self) -> str:return f"dim={self.dim}, input_resolution={self.input_resolution}, num_heads={self.num_heads}, " \f"window_size={self.window_size}, shift_size={self.shift_size}, mlp_ratio={self.mlp_ratio}"def flops(self):flops = 0H, W = self.input_resolution# norm1flops += self.dim * H * W# W-MSA/SW-MSAnW = H * W / self.window_size / self.window_sizeflops += nW * self.attn.flops(self.window_size * self.window_size)# mlpflops += 2 * H * W * self.dim * self.dim * self.mlp_ratio# norm2flops += self.dim * H * Wreturn flops#--------------------------#
# PatchMerging这个类的定义
#--------------------------#
class PatchMerging(nn.Module):r""" Patch Merging Layer.Args:input_resolution (tuple[int]): Resolution of input feature.dim (int): Number of input channels.norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm"""def __init__(self, input_resolution, dim, norm_layer=nn.LayerNorm):super(PatchMerging,self).__init__()self.input_resolution = input_resolutionself.dim = dimself.reduction = nn.Linear(4 * dim, 2 * dim, bias=False)self.norm = norm_layer(4 * dim)def forward(self, x):#-------------------## x的shape：B,H*W,C#-------------------#H, W = self.input_resolutionB, L, C = x.shapeassert L == H * W, "input feature has wrong size"assert H % 2 == 0 and W % 2 == 0, f"x size ({H}*{W}) are not even."x = x.view(B, H, W, C)x0 = x[:, 0::2, 0::2, :] # B H/2 W/2 Cx1 = x[:, 1::2, 0::2, :] # B H/2 W/2 Cx2 = x[:, 0::2, 1::2, :] # B H/2 W/2 Cx3 = x[:, 1::2, 1::2, :] # B H/2 W/2 C#-------------------------------## 在最后一个维度将四个tensor进行拼接# tensor的shape为 B H/2 W/2 4*C# 之后进行view操作即为reshape# 此时张量的shape为B H/2*W/2 4*C# 之后过LayerNorm# 最后通过全连接层来降低通道数# 此时的shape为B H/2*W/2 2*C#-------------------------------#x = torch.cat([x0, x1, x2, x3], -1)x = x.view(B, -1, 4 * C)x = self.norm(x)x = self.reduction(x)return x#-------------------------------## 定义extra_repr以及flops这两个函数#-------------------------------#def extra_repr(self) -> str:return f"input_resolution={self.input_resolution}, dim={self.dim}"def flops(self):H, W = self.input_resolutionflops = H * W * self.dimflops += (H // 2) * (W // 2) * 4 * self.dim * 2 * self.dimreturn flops#----------------------------------------------------#
# SwinTransformer整体结构中共有四个stage
# 四个stage中的layer的数目分别为2 2 6 2
# 这个类即为对于一个stage中所用到的layer的定义
#----------------------------------------------------#
class BasicLayer(nn.Module):r""" A basic Swin Transformer layer for one stage.Args:dim (int): Number of input channels.input_resolution (tuple[int]): Input resolution.depth (int): Number of blocks.num_heads (int): Number of attention heads.window_size (int): Local window size.mlp_ratio (float): Ratio of mlp hidden dim to embedding dim.qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: Trueqk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set.drop (float, optional): Dropout rate. Default: 0.0attn_drop (float, optional): Attention dropout rate. Default: 0.0drop_path (float | tuple[float], optional): Stochastic depth rate. Default: 0.0norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNormdownsample (nn.Module | None, optional): Downsample layer at the end of the layer. Default: Noneuse_checkpoint (bool): Whether to use checkpointing to save memory. Default: False. 是否使用checkpointing来节省内存，默认值为False"""def __init__(self, dim, input_resolution, depth, num_heads, window_size,mlp_ratio=4., qkv_bias=True, qk_scale=None, drop=0., attn_drop=0.,drop_path=0., norm_layer=nn.LayerNorm, downsample=None, use_checkpoint=False):super(BasicLayer,self).__init__()self.dim = dimself.input_resolution = input_resolutionself.depth = depthself.use_checkpoint = use_checkpoint#-----------------------------------## build blocks# 根据depth的值来确定一个layer中所用到的# SwinTransformerBlock的块数#-----------------------------------#self.blocks = nn.ModuleList([SwinTransformerBlock(dim=dim, input_resolution=input_resolution,num_heads=num_heads, window_size=window_size,shift_size=0 if (i % 2 == 0) else window_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)for i in range(depth)])#-----------------------------------------------------## patch merging layer# 在这里我们定义的downsample的值为None所以无downsample操作#-----------------------------------------------------#if downsample is not None:self.downsample = downsample(input_resolution, dim=dim, norm_layer=norm_layer)else:self.downsample = Nonedef forward(self, x):for blk in self.blocks:if self.use_checkpoint:x = checkpoint.checkpoint(blk, x)else:x = blk(x)if self.downsample is not None:x = self.downsample(x)return xdef extra_repr(self) -> str:return f"dim={self.dim}, input_resolution={self.input_resolution}, depth={self.depth}"def flops(self):flops = 0for blk in self.blocks:flops += blk.flops()if self.downsample is not None:flops += self.downsample.flops()return flops#------------------------------#
# PatchEmbed这个类的定义
#------------------------------#
class PatchEmbed(nn.Module):r""" Image to Patch EmbeddingArgs:img_size (int): Image size. Default: 224.patch_size (int): Patch token size. Default: 4.in_chans (int): Number of input image channels. Default: 3.embed_dim (int): Number of linear projection output channels. Default: 96.norm_layer (nn.Module, optional): Normalization layer. Default: None"""def __init__(self, img_size=224, patch_size=4, in_chans=3, embed_dim=96, norm_layer=None):super(PatchEmbed,self).__init__()#--------------------------------## to_2tuple()的作用在于生成一个元组# 并且该元组中有两个值相同的元素#--------------------------------#img_size = to_2tuple(img_size)patch_size = to_2tuple(patch_size)patches_resolution = [img_size[0] // patch_size[0], img_size[1] // patch_size[1]]self.img_size = img_sizeself.patch_size = patch_sizeself.patches_resolution = patches_resolutionself.num_patches = patches_resolution[0] * patches_resolution[1]self.in_chans = in_chansself.embed_dim = embed_dimself.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_size, stride=patch_size)if norm_layer is not None:self.norm = norm_layer(embed_dim)else:self.norm = Nonedef forward(self, x):B, C, H, W = x.shape# FIXME look at relaxing size constraintsassert H == self.img_size[0] and W == self.img_size[1], \f"Input image size ({H}*{W}) doesn't match model ({self.img_size[0]}*{self.img_size[1]})."#-----------------------------------------------## 1.先进行一次卷积,此时shape=batch_size, C, Ph, Pw# 2.将后两个维度进行flatten，使其成为Ph*Pw# 3.进行转置,此时的shape为batch_size,Ph*Pw C#-----------------------------------------------#x = self.proj(x).flatten(2).transpose(1, 2)if self.norm is not None:x = self.norm(x)return xdef flops(self):Ho, Wo = self.patches_resolutionflops = Ho * Wo * self.embed_dim * self.in_chans * (self.patch_size[0] * self.patch_size[1])if self.norm is not None:flops += Ho * Wo * self.embed_dimreturn flops#----------------------------------------------#
# The build of SwinTransformer
# 若在一个类中想要调用另外一个类,则我们并不
# 需要定义self方法,我们使用该类的类名直接进行调用即可
#----------------------------------------------#
class SwinTransformer(nn.Module):r""" Swin TransformerA PyTorch impl of : `Swin Transformer: Hierarchical Vision Transformer using Shifted Windows` -https://arxiv.org/pdf/2103.14030Args:img_size (int | tuple(int)): Input image size. Default 224patch_size (int | tuple(int)): Patch size. Default: 4in_chans (int): Number of input image channels. Default: 3num_classes (int): Number of classes for classification head. Default: 1000 -> ImageNetembed_dim (int): Patch embedding dimension. Default: 96depths (tuple(int)): Depth of each Swin Transformer layer.num_heads (tuple(int)): Number of attention heads in different layers.window_size (int): Window size. Default: 7mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. Default: 4qkv_bias (bool): If True, add a learnable bias to query, key, value. Default: Trueqk_scale (float): Override default qk scale of head_dim ** -0.5 if set. Default: Nonedrop_rate (float): Dropout rate. Default: 0attn_drop_rate (float): Attention dropout rate. Default: 0drop_path_rate (float): Stochastic depth rate. Default: 0.1norm_layer (nn.Module): Normalization layer. Default: nn.LayerNorm.ape (bool): If True, add absolute position embedding to the patch embedding. Default: Falsepatch_norm (bool): If True, add normalization after patch embedding. Default: Trueuse_checkpoint (bool): Whether to use checkpointing to save memory. Default: False"""def __init__(self, img_size=224, patch_size=4, in_chans=3, num_classes=1000,embed_dim=96, depths=[2, 2, 6, 2], num_heads=[3, 6, 12, 24],window_size=7, 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, ape=False, patch_norm=True,use_checkpoint=False, **kwargs):super(SwinTransformer,self).__init__()self.num_classes = num_classesself.num_layers = len(depths)self.embed_dim = embed_dimself.ape = apeself.patch_norm = patch_normself.num_features = int(embed_dim * 2 ** (self.num_layers - 1))self.mlp_ratio = mlp_ratio#------------------------------------------------## split image into non-overlapping patches# 将输入的图片split成多个patch#------------------------------------------------#self.patch_embed = PatchEmbed(img_size = img_size, patch_size=patch_size, in_chans=in_chans, embed_dim=embed_dim,norm_layer = norm_layer if self.patch_norm else None)num_patches = self.patch_embed.num_patchespatches_resolution = self.patch_embed.patches_resolutionself.patches_resolution = patches_resolution#---------------------------------------## absolute position embedding# 类中定义的ape=False,所以我们直接pass即可#---------------------------------------#if self.ape:self.absolute_pos_embed = nn.Parameter(torch.zeros(1, num_patches, embed_dim))trunc_normal_(self.absolute_pos_embed, std=.02)#----------------------------## 定义一个Dropout层#----------------------------#self.pos_drop = nn.Dropout(p=drop_rate)# stochastic depthdpr = [x.item() for x in torch.linspace(0, drop_path_rate, sum(depths))] # stochastic depth decay rule#--------------------------------------------------## build layers# 首先定义一个空的ModuleList# 之后根据layer的数目将BasicLayer添加至ModuleList中#--------------------------------------------------#self.layers = nn.ModuleList()for i_layer in range(self.num_layers):layer = BasicLayer(dim=int(embed_dim * 2 ** i_layer),input_resolution=(patches_resolution[0] // (2 ** i_layer),patches_resolution[1] // (2 ** i_layer)),depth = depths[i_layer],num_heads = num_heads[i_layer],window_size = window_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[sum(depths[:i_layer]):sum(depths[:i_layer + 1])],norm_layer = norm_layer,downsample = PatchMerging if (i_layer < self.num_layers - 1) else None,use_checkpoint = use_checkpoint)self.layers.append(layer)#-------------------------------------## 定义LayerNorm层# 定义自适应二维平均池化层# 定义head,使用Linear来实现# 之后进行了一个网络权重初始化的定义#-------------------------------------#self.norm = norm_layer(self.num_features)self.avgpool = nn.AdaptiveAvgPool1d(1)self.head = nn.Linear(self.num_features, num_classes) if num_classes > 0 else nn.Identity()self.apply(self._init_weights)#-------------------------------## 网络的权重初始化#-------------------------------#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.ignoredef no_weight_decay(self):return {'absolute_pos_embed'}@torch.jit.ignoredef no_weight_decay_keywords(self):return {'relative_position_bias_table'}#--------------------------------------------------------------------------------------## forward_features函数的定义# 首先将输入的图片打成多个小的patch# 之后过Dropout来降低发生过拟合的风险,因为Linear层的存在所以说会定义比较多层的Dropout# 过SwinTransformer的四个stage,过LayerNorm,再过二维平均池化层,最后将后面的两个维度进行flatten#--------------------------------------------------------------------------------------#def forward_features(self, x):x = self.patch_embed(x)if self.ape:x = x + self.absolute_pos_embedx = self.pos_drop(x)for layer in self.layers:x = layer(x)#-----------------------------## shape的变化# B L C -> B C 1 -> B C#-----------------------------#x = self.norm(x)x = self.avgpool(x.transpose(1, 2))x = torch.flatten(x, 1)return x#---------------------------------## 主要为运用Linear层将特征图像的# 通道数转变为ImageNet上所规定的类别数#---------------------------------#def forward(self, x):x = self.forward_features(x)x = self.head(x)return xdef flops(self):flops = 0flops += self.patch_embed.flops()for i, layer in enumerate(self.layers):flops += layer.flops()flops += self.num_features * self.patches_resolution[0] * self.patches_resolution[1] // (2 ** self.num_layers)flops += self.num_features * self.num_classesreturn flops
Swin-Transformer通用视觉骨干网络主体结构代码解释相关推荐
- Swin Transformer（W-MSA详解）代码+图解
2. Window & Shifted Window based Self-Attention Swin Transformer另一个重要的改进就是window-based的self-atte ...
- 重磅开源！屠榜各大CV任务！最强骨干网络：Swin Transformer来了
点击下方卡片,关注"CVer"公众号 AI/CV重磅干货,第一时间送达 本文转载自:Smarter Swin Transformer 代码于2021年4月13日凌晨刚刚开源! ht ...
- 【CV】Swin Transformer：使用 Shifted Windows 的分层视觉 Transformer
论文名称:Swin Transformer: Hierarchical Vision Transformer using Shifted Windows 论文下载:https://arxiv.org/ ...
- 【读点论文】A ConvNet for the 2020s，结合swin transformer的结构设计和训练技巧调整resnet网络，在类似的FLOPs和参数量取得更好一点的效果
A ConvNet for the 2020s Abstract 视觉识别的"咆哮的20年代"始于视觉transformer(ViTs)的问世,它迅速取代ConvNets成为最先进 ...
- VisualTransformer（ViT）视觉任务骨干网络
在这里分享一下我对于ViT骨干网络的代码理解,ViT paper发表于2020年,掀起了transformer结构在视觉任务中的应用潮流. paper:https://arxiv.org/abs/20 ...
- Swin Transformer原理与代码精讲
课程链接:Swin Transformer原理与代码精讲--计算机视觉视频教程-人工智能-CSDN程序员研修院 Transformer在许多NLP(自然语言处理)任务中取得了最先进的成果. Swin ...
- ICCV 2021 Best Paper | Swin Transformer何以屠榜各大CV任务！
作者:陀飞轮@知乎(已授权) 来源:https://zhuanlan.zhihu.com/p/360513527 编辑:智源社区 近日,Swin Transformer拿到2021 ICCV Best ...
- YOLOv5/v7 更换骨干网络之 SwinTransformer
提供 YOLOv5 / YOLOv7 / YOLOv7-tiny 模型 YAML 文件 论文地址:https://arxiv.org/pdf/2103.14030.pdf 代码地址:https://g ...
- 【深度学习】论文阅读：（ICCV-2021））Swin Transformer
这里写目录标题 论文详情 VIT缺点 改进点 概述 核心思想 整体结构 名称解释 Window.Patch.Token 与vit区别 结构过程 Patch Embedding BasicLayer P ...
最新文章
- HDU4549 M斐波那契数列(矩阵快速幂+费马小定理)
- 009_InputNumber计数器
- 机器学习理论《统计学习方法》学习笔记：第九章 EM算法及其推广
- Eclipse Add generated serial version ID报错解决方案
- HDP 2.6 requires libtirpc-devel
- 接口测试从零开始系列_mock技术使用
- Python描述符是什么？
- 【LeetCode】【HOT】98. 验证二叉搜索树（递归）
- ASP.Net中服务器控件的生命周期
- 以太坊web3.js文档翻译及说明
- DotNetNuke 04.05.05 安装
- 1015 德才论 (25 分)—PAT (Basic Level) Practice （中文）
- pandas学习-task2
- Linux从入门到精通二（Windows:你知道我的兄弟Linux吗？）
- perl linux yum,Linux CentOS6.5(x86_64)安装Perl5.26
- 微星主板开启安全启动以更新win11教程
- AJAX简单思维导图
- speedoffice表格中怎样给文字添加删除线？
- 插件自动解决谷歌翻译用不了，win、mac、linux通用，附链接
- mycat分库分表（线上实际案例）