Swin Transformer
论文名称： Swin Transformer: Hierarchical Vision Transformer using Shifted Windows
一、PatchEmbed（对应结构图中patch partition和linear embedding）
class PatchEmbed(nn.Module): # 将图片划分成一个个不重叠的patches
"""
2D Image to Patch Embedding
"""
def __init__(self, patch_size=4, in_c=3, embed_dim=96, norm_layer=None):
super().__init__()
patch_size = (patch_size, patch_size)
self.patch_size = patch_size
self.in_chans = in_c
self.embed_dim = embed_dim
self.proj = nn.Conv2d(in_c, embed_dim, kernel_size=patch_size, stride=patch_size) # 对应Patch Partition和Linrer Embedding层
self.norm = norm_layer(embed_dim) if norm_layer else nn.Identity() #判断是否norm_layer
def forward(self, x):
_, _, H, W = x.shape
# padding
# 如果输入图片的H，W不是patch_size的整数倍，需要进行padding
pad_input = (H % self.patch_size[0] != 0) or (W % self.patch_size[1] != 0)
if pad_input:
# to pad the last 3 dimensions(C, H, W), 从后向前进行，即W->H->C
# (W_left, W_right, H_top,H_bottom, C_front, C_back)
x = F.pad(x, (0, self.patch_size[1] - W % self.patch_size[1],
0, self.patch_size[0] - H % self.patch_size[0],
0, 0)) #在宽的右侧，高的下侧进行填充0
# 下采样patch_size倍
x = self.proj(x) #[B, 3, H, W] -> [B, 96, H/4, W/4]
_, _, H, W = x.shape
# flatten: [B, C, H, W] -> [B, C, HW]
# transpose: [B, C, HW] -> [B, HW, C]
x = x.flatten(2).transpose(1, 2)
x = self.norm(x)
return x, H, W二、Patch Merging
class PatchMerging(nn.Module): # 定义Patch Merging层
def __init__(self, dim, norm_layer=nn.LayerNorm):
super().__init__()
self.dim = dim
self.reduction = nn.Linear(4 * dim, 2 * dim, bias=False)
self.norm = norm_layer(4 * dim)
def forward(self, x, H, W): # 正向传播
"""
x: B, H*W, C
"""
B, L, C = x.shape
assert L == H * W, "input feature has wrong size"
x = x.view(B, H, W, C)
# padding
# 如果输入feature map的H，W不是2的整数倍，需要进行padding
pad_input = (H % 2 == 1) or (W % 2 == 1)
if pad_input:
# to pad the last 3 dimensions(H, W, C), starting from the last dimension and moving forward, 即C->W->H.
# (C_front, C_back, W_left, W_right, H_top, H_bottom)
# 注意这里的Tensor通道是[B, H, W, C]，所以会和官方文档有些不同
x = F.pad(x, (0, 0, 0, W % 2, 0, H % 2)) #在宽的右侧，高的下侧进行填充0
x0 = x[:, 0::2, 0::2, :] # [B, H/2, W/2, C] 宽高方向上都从0开始，以2为步长采样，得到上图蓝色像素点
x1 = x[:, 1::2, 0::2, :] # [B, H/2, W/2, C] 宽方向上从0开始，高方向上从1开始，以2为步长采样，得到上图绿色像素点
x2 = x[:, 0::2, 1::2, :] # [B, H/2, W/2, C] 宽方向上从1开始，高方向上从0开始，以2为步长采样，得到上图黄色像素点
x3 = x[:, 1::2, 1::2, :] # [B, H/2, W/2, C] 宽高方向上都从1开始，以2为步长采样，得到上图红色像素点
x = torch.cat([x0, x1, x2, x3], -1) # [B, H/2, W/2, 4*C] 在通道维度上进行拼接
x = x.view(B, -1, 4 * C) # [B, H/2*W/2, 4*C]
x = self.norm(x) #通过norm_layer层
x = self.reduction(x) # [B, H/2*W/2, 2*C] 通过全连接层
return x三、MLP模块（与Vision Transformer中的相同）
class Mlp(nn.Module): # 定义Mlp模块
""" MLP as used in Vision Transformer, MLP-Mixer and related networks
"""
def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=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.drop1 = nn.Dropout(drop)
self.fc2 = nn.Linear(hidden_features, out_features)
self.drop2 = nn.Dropout(drop)
def forward(self, x):
x = self.fc1(x)
x = self.act(x)
x = self.drop1(x)
x = self.fc2(x)
x = self.drop2(x)
return x四、Swin Transformer Block模块
class SwinTransformerBlock(nn.Module):
def __init__(self, dim, num_heads, window_size=7, shift_size=0,
mlp_ratio=4., qkv_bias=True, drop=0., attn_drop=0., drop_path=0.,
act_layer=nn.GELU, norm_layer=nn.LayerNorm):
super().__init__()
self.dim = dim
self.num_heads = num_heads
self.window_size = window_size
self.shift_size = shift_size
self.mlp_ratio = mlp_ratio
assert 0 <= self.shift_size < self.window_size, "shift_size must in 0-window_size"
self.norm1 = norm_layer(dim)
self.attn = WindowAttention(
dim, window_size=(self.window_size, self.window_size), num_heads=num_heads, qkv_bias=qkv_bias,
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)
def forward(self, x, attn_mask):
H, W = self.H, self.W
B, L, C = x.shape
assert L == H * W, "input feature has wrong size"
shortcut = x
x = self.norm1(x)
x = x.view(B, H, W, C) #[B, L, C] -> [B, H, W, C]
# pad feature maps to multiples of window size
# 把feature map给pad到window size的整数倍
pad_l = pad_t = 0
pad_r = (self.window_size - W % self.window_size) % self.window_size
pad_b = (self.window_size - H % self.window_size) % self.window_size
x = F.pad(x, (0, 0, pad_l, pad_r, pad_t, pad_b))
_, Hp, Wp, _ = x.shape #获取pad后的宽高
# cyclic shift
if self.shift_size > 0: #为SW-MSA
# dims=(1, 2)对应(B, H, W, C)中 H,W，从上往下移shift_size行，从左移往右shift_size列
shifted_x = torch.roll(x, shifts=(-self.shift_size, -self.shift_size), dims=(1, 2))
else: #为W-MSA
shifted_x = x
attn_mask = None
# partition windows
x_windows = window_partition(shifted_x, self.window_size) # [nW*B, Mh, Mw, C] 通过window_partition方法划分窗口
x_windows = x_windows.view(-1, self.window_size * self.window_size, C) # [nW*B, Mh*Mw, C]
# W-MSA/SW-MSA
attn_windows = self.attn(x_windows, mask=attn_mask) # [nW*B, Mh*Mw, C]
# merge windows
attn_windows = attn_windows.view(-1, self.window_size, self.window_size, C) # [nW*B, Mh, Mw, C]
shifted_x = window_reverse(attn_windows, self.window_size, Hp, Wp) # [B, H', W', C] 将window还原成一个feature map
# reverse cyclic shift
if self.shift_size > 0: #为SW-MSA
# dims=(1, 2)对应(B, H, W, C)中 H,W，从下往上移shift_size行，从右往左移shift_size列
x = torch.roll(shifted_x, shifts=(self.shift_size, self.shift_size), dims=(1, 2))
else: #为W-MSA
x = shifted_x
if pad_r > 0 or pad_b > 0:
# 把前面pad的数据移除掉，通过contiguous方法变成内存中的连续数据
x = x[:, :H, :W, :].contiguous()
x = x.view(B, H * W, C) #[B, L, C]
# FFN
x = shortcut + self.drop_path(x)
x = x + self.drop_path(self.mlp(self.norm2(x)))
return xdef window_partition(x, window_size: int):
"""
将feature map按照window_size划分成一个个没有重叠的window
Args:
x: (B, H, W, C)
window_size (int): window size(M)
Returns:
windows: (num_windows*B, window_size, window_size, C)
"""
B, H, W, C = x.shape
x = x.view(B, H // window_size, window_size, W // window_size, window_size, C)
# permute: [B, H//Mh, Mh, W//Mw, Mw, C] -> [B, H//Mh, W//Mh, Mw, Mw, C]
# view: [B, H//Mh, W//Mw, Mh, Mw, C] -> [B*num_windows, Mh, Mw, C]
windows = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(-1, window_size, window_size, C) #contiguous方法变成内存连续形式
return windows
def window_reverse(windows, window_size: int, H: int, W: int):
"""
将一个个window还原成一个feature map
Args:
windows: (num_windows*B, window_size, window_size, C)
window_size (int): Window size(M)
H (int): Height of image
W (int): Width of image
Returns:
x: (B, H, W, C)
"""
B = int(windows.shape[0] / (H * W / window_size / window_size))
# view: [B*num_windows, Mh, Mw, C] -> [B, H//Mh, W//Mw, Mh, Mw, C]
x = windows.view(B, H // window_size, W // window_size, window_size, window_size, -1)
# permute: [B, H//Mh, W//Mw, Mh, Mw, C] -> [B, H//Mh, Mh, W//Mw, Mw, C]
# view: [B, H//Mh, Mh, W//Mw, Mw, C] -> [B, H, W, C]
x = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(B, H, W, -1) #contiguous方法变成内存连续形式
return x五、W-MSA/SW-MSA模块
class WindowAttention(nn.Module):
def __init__(self, dim, window_size, num_heads, qkv_bias=True, attn_drop=0., proj_drop=0.):
super().__init__()
self.dim = dim
self.window_size = window_size # [Mh, Mw]
self.num_heads = num_heads
head_dim = dim // num_heads
self.scale = head_dim ** -0.5
# define a parameter table of relative position bias
self.relative_position_bias_table = nn.Parameter(
torch.zeros((2 * window_size[0] - 1) * (2 * window_size[1] - 1), num_heads)) # [2*Mh-1 * 2*Mw-1, nH]
# get pair-wise relative position index for each token inside the window
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])) # [2, Mh, Mw]
coords_flatten = torch.flatten(coords, 1) # [2, Mh*Mw]
# [2, Mh*Mw, 1] - [2, 1, Mh*Mw]
relative_coords = coords_flatten[:, :, None] - coords_flatten[:, None, :] # [2, Mh*Mw, Mh*Mw]
relative_coords = relative_coords.permute(1, 2, 0).contiguous() # [Mh*Mw, Mh*Mw, 2]
relative_coords[:, :, 0] += self.window_size[0] - 1 # shift to start from 0
relative_coords[:, :, 1] += self.window_size[1] - 1
relative_coords[:, :, 0] *= 2 * self.window_size[1] - 1
relative_position_index = relative_coords.sum(-1) # [Mh*Mw, Mh*Mw]
self.register_buffer("relative_position_index", relative_position_index)
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)
nn.init.trunc_normal_(self.relative_position_bias_table, std=.02)
self.softmax = nn.Softmax(dim=-1)
def forward(self, x, mask: Optional[torch.Tensor] = None):
"""
Args:
x: input features with shape of (num_windows*B, Mh*Mw, C)
mask: (0/-inf) mask with shape of (num_windows, Wh*Ww, Wh*Ww) or None
"""
# [batch_size*num_windows, Mh*Mw, total_embed_dim]
B_, N, C = x.shape
# qkv(): -> [batch_size*num_windows, Mh*Mw, 3 * total_embed_dim]
# reshape: -> [batch_size*num_windows, Mh*Mw, 3, num_heads, embed_dim_per_head]
# permute: -> [3, batch_size*num_windows, num_heads, Mh*Mw, embed_dim_per_head]
qkv = self.qkv(x).reshape(B_, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)
# [batch_size*num_windows, num_heads, Mh*Mw, embed_dim_per_head]
q, k, v = qkv.unbind(0) # make torchscript happy (cannot use tensor as tuple)
# transpose: -> [batch_size*num_windows, num_heads, embed_dim_per_head, Mh*Mw]
# @: multiply -> [batch_size*num_windows, num_heads, Mh*Mw, Mh*Mw]
q = q * self.scale
attn = (q @ k.transpose(-2, -1))
# relative_position_bias_table.view: [Mh*Mw*Mh*Mw,nH] -> [Mh*Mw,Mh*Mw,nH]
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() # [nH, Mh*Mw, Mh*Mw]
attn = attn + relative_position_bias.unsqueeze(0)
if mask is not None:
# mask: [nW, Mh*Mw, Mh*Mw]
nW = mask.shape[0] # num_windows
# attn.view: [batch_size, num_windows, num_heads, Mh*Mw, Mh*Mw]
# mask.unsqueeze: [1, nW, 1, Mh*Mw, Mh*Mw]
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)
# @: multiply -> [batch_size*num_windows, num_heads, Mh*Mw, embed_dim_per_head]
# transpose: -> [batch_size*num_windows, Mh*Mw, num_heads, embed_dim_per_head]
# reshape: -> [batch_size*num_windows, Mh*Mw, total_embed_dim]
x = (attn @ v).transpose(1, 2).reshape(B_, N, C)
x = self.proj(x)
x = self.proj_drop(x)
return x六、stage模块
class BasicLayer(nn.Module): # 定义stage模块
def __init__(self, dim, depth, num_heads, window_size,
mlp_ratio=4., qkv_bias=True, drop=0., attn_drop=0.,
drop_path=0., norm_layer=nn.LayerNorm, downsample=None, use_checkpoint=False):
super().__init__()
self.dim = dim
self.depth = depth
self.window_size = window_size
self.use_checkpoint = use_checkpoint
self.shift_size = window_size // 2 #SW-MSA中将窗口向右向下移动的距离,window_size//2向下取整
# build blocks
self.blocks = nn.ModuleList([
SwinTransformerBlock(
dim=dim,
num_heads=num_heads,
window_size=window_size,
shift_size=0 if (i % 2 == 0) else self.shift_size, #判断使用的是W-MSA(shift_size=0)还是SW-MSA(shift_size)
mlp_ratio=mlp_ratio,
qkv_bias=qkv_bias,
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 判断stage模块中是否有patch merging层
if downsample is not None:
self.downsample = downsample(dim=dim, norm_layer=norm_layer)
else:
self.downsample = None
def create_mask(self, x, H, W):
# calculate attention mask for SW-MSA
# 保证Hp和Wp是window_size的整数倍
Hp = int(np.ceil(H / self.window_size)) * self.window_size
Wp = int(np.ceil(W / self.window_size)) * self.window_size
# 拥有和feature map一样的通道排列顺序，方便后续window_partition
img_mask = torch.zeros((1, Hp, Wp, 1), device=x.device) # [1, Hp, Wp, 1]
h_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
for h in h_slices:
for w in w_slices:
img_mask[:, h, w, :] = cnt # [B, H, W, C]
cnt += 1
mask_windows = window_partition(img_mask, self.window_size) # [nW, Mh, Mw, 1]
mask_windows = mask_windows.view(-1, self.window_size * self.window_size) # [nW, Mh*Mw]
attn_mask = mask_windows.unsqueeze(1) - mask_windows.unsqueeze(2) # [nW, 1, Mh*Mw] - [nW, Mh*Mw, 1]
# [nW, Mh*Mw, Mh*Mw]
attn_mask = attn_mask.masked_fill(attn_mask != 0, float(-100.0)).masked_fill(attn_mask == 0, float(0.0))
return attn_mask
def forward(self, x, H, W):
attn_mask = self.create_mask(x, H, W) # [nW, Mh*Mw, Mh*Mw]
for blk in self.blocks:
blk.H, blk.W = H, W
if not torch.jit.is_scripting() and self.use_checkpoint:
x = checkpoint.checkpoint(blk, x, attn_mask)
else:
x = blk(x, attn_mask)
if self.downsample is not None:
x = self.downsample(x, H, W)
H, W = (H + 1) // 2, (W + 1) // 2
return x, H, W6.1、SW-MSA中mask创建
def create_mask(self, x, H, W):
# calculate attention mask for SW-MSA
# 保证Hp和Wp是window_size的整数倍
Hp = int(np.ceil(H / self.window_size)) * self.window_size
Wp = int(np.ceil(W / self.window_size)) * self.window_size
# 拥有和feature map一样的通道排列顺序，方便后续window_partition
img_mask = torch.zeros((1, Hp, Wp, 1), device=x.device) # [1, Hp, Wp, 1]
h_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
for h in h_slices:
for w in w_slices:
img_mask[:, h, w, :] = cnt # [B, H, W, C]
cnt += 1七、Swin Transformer整体结构
class SwinTransformer(nn.Module):
def __init__(self, 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,
drop_rate=0., attn_drop_rate=0., drop_path_rate=0.1,
norm_layer=nn.LayerNorm, patch_norm=True,
use_checkpoint=False, **kwargs):
super().__init__()
self.num_classes = num_classes
self.num_layers = len(depths)
self.embed_dim = embed_dim
self.patch_norm = patch_norm
# stage4输出特征矩阵的channels
self.num_features = int(embed_dim * 2 ** (self.num_layers - 1)) # embed_dim * 8
self.mlp_ratio = mlp_ratio
# 将图片划分成一个个不重叠的patches (split image into non-overlapping patches)
self.patch_embed = PatchEmbed(
patch_size=patch_size, in_c=in_chans, embed_dim=embed_dim,
norm_layer=norm_layer if self.patch_norm else None)
self.pos_drop = nn.Dropout(p=drop_rate)
# stochastic depth 创建递增的dpr
dpr = [x.item() for x in torch.linspace(0, drop_path_rate, sum(depths))] # stochastic depth decay rule
# build layers
self.layers = nn.ModuleList()
for i_layer in range(self.num_layers): #self.num_layers=4
# 注意这里构建的stage和论文图中有些差异
# 这里的stage不包含该stage的patch_merging层，包含的是下个stage的
layers = BasicLayer(dim=int(embed_dim * 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,
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, # Patch Merging层，只有前三个stage有
use_checkpoint=use_checkpoint)
self.layers.append(layers)
# 分类器
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)
def _init_weights(self, m): # 权重初始化
if isinstance(m, nn.Linear):
nn.init.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)
def forward(self, x): # 定义正向传播过程
# x: [B, L, C]
x, H, W = self.patch_embed(x) # 先通过patch_embed层
x = self.pos_drop(x) # 通过dropout层
for layer in self.layers: # 依次通过stage1、stage2、stage3、stage4
x, H, W = layer(x, H, W)
x = self.norm(x) # [B, L, C]
x = self.avgpool(x.transpose(1, 2)) # [B, C, 1]
x = torch.flatten(x, 1)
x = self.head(x)
return x
reference
【12.2 使用Pytorch搭建Swin-Transformer网络】 https://www.bilibili.com/video/BV1yg411K7Yc/?share_source=copy_web&vd_source=95705b32f23f70b32dfa1721628d5874