【导读】
当ViT因高分辨率图像计算效率低下而受阻，当CNN难以捕捉全局依赖关系------2021年横空出世的Swin Transformer，以"移位窗口注意力"和分层设计，在ImageNet、COCO等基准任务中刷新记录。本文将深入解析其革命性架构，并附PyTorch实战代码！
添加图片注释，不超过 140 字（可选）
一、为什么需要Swin Transformer？
- ViT的瓶颈：全局注意力计算复杂度随图像分辨率呈平方级增长（O(n²)），难以处理高分辨率图像
- CNN的局限：卷积核的局部性限制全局上下文建模能力
Swin Transformers（移位窗口 Transformers）解决了 ViT 的低效问题，由于二次方注意力机制的复杂性，ViT 难以处理高分辨率图像。Swin 引入了基于局部窗口的注意力机制和多尺度层次结构，将类似 CNN 的局部性与 Transformer 的灵活性相结合。
添加图片注释，不超过 140 字（可选）
Swin的突破：
- 线性计算复杂度（O(n)）
- 融合CNN的局部性优势与Transformer的全局建模能力
类比：如同望远镜，既能聚焦局部细节（窗口内注意力），又能通过移动视野建立全局关联（移位窗口）
二、Swin Transformers的主要特点
移位窗口注意力（Shifted Window）
与 ViT 不同，ViT 会计算所有图像块的注意力（对于大图像来说成本很高）：
- 将图像分成不重叠的窗口（例如，7×7 块）。
- 在每个窗口内应用自我注意力，将复杂性从二次方降低到线性。
- 交替使用常规窗口和移位窗口来实现跨窗口通信。
分层金字塔结构
Swin 分阶段处理图像，模仿 CNN 池化：
- 补丁嵌入：将 4×4 图像补丁转换为标记。
- 局部注意力块：应用基于窗口的多头自注意力（W-MSA）。
- 补丁合并：组合 2×2 补丁以降低分辨率，增加感受视野。
Swin 变压器块
每个区块包括：
- LayerNorm 用于稳定训练。
- 基于窗口的多头自注意力（W-MSA）或移位 W-MSA（SW-MSA）。
- 用于非线性变换的 MLP（前馈）。
- 残差连接可防止梯度消失。
添加图片注释，不超过 140 字（可选）
三、为什么Swin的表现优于ViT和CNN
- 效率：线性复杂度（O（n））与 ViT 的二次方（O（n²））相比，非常适合高分辨率图像。
- 可扩展性：Swin-V2 可处理十亿参数模型和 1536×1536 图像，并进行连续相对位置偏差等调整。
- 性能：在ImageNet-1K上实现87.3%的top-1 准确率，在COCO检测上实现58.7的box AP，在ADE20K分割上实现 53.5 mIoU。
- 多功能性：分类、检测、分割和医学成像的通用主干。
四、PyTorch 实现 Swin 概述
Swin Transformer 类
初始化参数：-在各种其他 dropout 和 normalization 参数中，这些参数包括：window_size局部自注意的窗口大小。ape (bool)：如果为 True，则将绝对位置嵌入添加到补丁嵌入。fused_window_process：可选的硬件优化。
应用补丁嵌入：与 ViT 类似，图像被分成不重叠的补丁并使用线性嵌入Conv2D。
应用位置嵌入：SwinTransformer可选地使用绝对位置嵌入 (ape)，并将其添加到图像块嵌入中。绝对位置嵌入通常有助于模型学习利用每个图像块的位置信息来做出更明智的预测。
应用深度衰减：深度衰减有助于正则化并防止过拟合。深度衰减通常通过在训练期间跳过某些层来实现。在此 Swin 实现中，我们采用了随机深度衰减，这意味着层越深，被跳过的可能性就越高。
层级构建：该模型由多层（BasicLayer）构成SwinTransformerBlock，每层使用 对特征图进行下采样，以便进行分层处理PatchMerging。特征的维度和特征图的分辨率会随着层级的变化而变化。
分类头：与 ViT 类似，它使用多层感知器 (MLP) 头进行分类任务，如self.head最后一步中所定义。
ini
lass SwinTransformer(nn.Module):
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=4.0,
qkv_bias=True,
qk_scale=None,
drop_rate=0.0,
attn_drop_rate=0.0,
drop_path_rate=0.1,
norm_layer=nn.LayerNorm,
ape=False,
patch_norm=True,
use_checkpoint=False,
fused_window_process=False,
**kwargs,
):
super().__init__()
self.num_classes = num_classes
self.num_layers = len(depths)
self.embed_dim = embed_dim
self.ape = ape
self.patch_norm = patch_norm
self.num_features = int(embed_dim * 2 ** (self.num_layers - 1))
self.mlp_ratio = mlp_ratio
# split image into non-overlapping patches
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_patches
patches_resolution = self.patch_embed.patches_resolution
self.patches_resolution = patches_resolution
# absolute position embedding
if self.ape:
self.absolute_pos_embed = nn.Parameter(torch.zeros(1, num_patches, embed_dim))
trunc_normal_(self.absolute_pos_embed, std=0.02)
self.pos_drop = nn.Dropout(p=drop_rate)
# stochastic depth
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):
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,
fused_window_process=fused_window_process,
)
self.layers.append(layer)
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):
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 forward_features(self, x):
x = self.patch_embed(x)
if self.ape:
x = x + self.absolute_pos_embed
x = self.pos_drop(x)
for layer in self.layers:
x = layer(x)
x = self.norm(x) # B L C
x = self.avgpool(x.transpose(1, 2)) # B C 1
x = torch.flatten(x, 1)
return x
def forward(self, x):
x = self.forward_features(x)
x = self.head(x)
return xSwin 变压器块
Swin Transformer的核心操作：局部窗口注意力机制和后续的 MLP 处理。它对于 Swin Transformer 能够通过关注局部块来高效处理大图像，同时保持学习全局表征的能力起着关键作用。
图层组件：
- 规范化层 1（self.norm1）：应用于注意力机制之前。
- 窗口注意力（self.attn）：计算局部窗口内的自我注意力。
- 丢弃路径（self.drop_path）：实现随机深度以进行正则化。
- 规范化层 2（self.norm2）：应用于 MLP 层之前。
- MLP（mlp）：用于处理后注意特征的多层感知器。
- 注意力掩码(self.register_buffer)：注意力掩码用于自注意力计算，以控制窗口输入中哪些元素可以交互（即相互关注）。移位窗口方法允许一些跨窗口交互，从而帮助模型捕获更广泛的上下文信息。
ini
class SwinTransformerBlock(nn.Module):
r"""Swin Transformer Block.
Args:
dim (int): Number of input channels.
input_resolution (tuple[int]): Input resulotion.
num_heads (int): Number of attention heads.
window_size (int): Window size.
shift_size (int): Shift size for SW-MSA.
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: True
qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set.
drop (float, optional): Dropout rate. Default: 0.0
attn_drop (float, optional): Attention dropout rate. Default: 0.0
drop_path (float, optional): Stochastic depth rate. Default: 0.0
act_layer (nn.Module, optional): Activation layer. Default: nn.GELU
norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm
fused_window_process (bool, optional): If True, use one kernel to fused window shift & window partition for acceleration, similar for the reversed part. Default: False
"""
def __init__(
self,
dim,
input_resolution,
num_heads,
window_size=7,
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,
fused_window_process=False,
):
super().__init__()
self.dim = dim
self.input_resolution = input_resolution
self.num_heads = num_heads
self.window_size = window_size
self.shift_size = shift_size
self.mlp_ratio = mlp_ratio
if min(self.input_resolution) <= self.window_size:
# if window size is larger than input resolution, we don't partition windows
self.shift_size = 0
self.window_size = min(self.input_resolution)
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=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 > 0.0 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
H, W = self.input_resolution
img_mask = torch.zeros((1, H, W, 1)) # 1 H W 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
cnt += 1
mask_windows = window_partition(
img_mask, self.window_size
) # nW, window_size, window_size, 1
mask_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 = None
self.register_buffer("attn_mask", attn_mask)
self.fused_window_process = fused_window_process
### New cell ###
def forward(self, x):
H, W = self.input_resolution
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)
# cyclic shift
if self.shift_size > 0:
if not self.fused_window_process:
shifted_x = torch.roll(
x, shifts=(-self.shift_size, -self.shift_size), dims=(1, 2)
)
# partition windows
x_windows = window_partition(
shifted_x, self.window_size
) # nW*B, window_size, window_size, C
else:
x_windows = WindowProcess.apply(
x, B, H, W, C, -self.shift_size, self.window_size
)
else:
shifted_x = x
# partition windows
x_windows = window_partition(
shifted_x, self.window_size
) # nW*B, window_size, window_size, C
x_windows = x_windows.view(
-1, self.window_size * self.window_size, C
) # nW*B, window_size*window_size, C
# W-MSA/SW-MSA
attn_windows = self.attn(
x_windows, mask=self.attn_mask
) # nW*B, window_size*window_size, C
# merge windows
attn_windows = attn_windows.view(-1, self.window_size, self.window_size, C)
# reverse cyclic shift
if self.shift_size > 0:
if not self.fused_window_process:
shifted_x = window_reverse(
attn_windows, self.window_size, H, W
) # B H' W' C
x = torch.roll(
shifted_x, shifts=(self.shift_size, self.shift_size), dims=(1, 2)
)
else:
x = WindowProcessReverse.apply(
attn_windows, B, H, W, C, self.shift_size, self.window_size
)
else:
shifted_x = window_reverse(attn_windows, self.window_size, H, W) # B H' W' C
x = shifted_x
x = x.view(B, H * W, C)
x = shortcut + self.drop_path(x)
# Feed-forward network (FFN)
x = x + self.drop_path(self.mlp(self.norm2(x)))
return xSwin Transformer Block 的前向传播
- 循环移位（cyclic shift）是在对特征图进行窗口划分之前应用于整个特征图的操作。 这是Swin Transformer实现"移位窗口"机制的关键一步。具体流程是：
- 常规层(l%2==0):直接将特征图划分 (window_partition) 成不重叠的 MxM 窗口，然后在每个窗口内计算自注意力 (W-MSA)。
- 移位层 (l % 2 != 0): 首先 将特征图在空间维度上 循环移位 (例如，左上角移出窗口的部分会出现在右下角)。然后 将移位后的特征图划分 (window_partition) 成 看似 与常规层相同的 MxM 窗口（但实际包含了来自原始位置相邻窗口的特征块）。在这些"新"窗口内计算自注意力 (SW-MSA)。最后，非常重要的一步是 将计算完注意力的窗口特征 反向循环移位 (reverse_cyclic_shift) 回原始位置，再合并 (window_reverse) 成特征图。
窗口注意
WindowAttention是一个基于窗口的多头自注意力 (W-MSA) 模块，具有相对位置偏差。它既可以用于移位窗口，也可以用于非移位窗口。
ini
class WindowAttention(nn.Module):
"""
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: True
qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set
attn_drop (float, optional): Dropout ratio of attention weight. Default: 0.0
proj_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=0.0,
proj_drop=0.0,
):
super().__init__()
self.dim = dim
self.window_size = window_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 * window_size[0] - 1) * (2 * window_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.window_size[0])
coords_w = torch.arange(self.window_size[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.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) # Wh*Ww, Wh*Ww
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)
trunc_normal_(self.relative_position_bias_table, std=0.02)
self.softmax = nn.Softmax(dim=-1)
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
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],
) # make torchscript happy (cannot use tensor as tuple)
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.window_size[0] * self.window_size[1],
self.window_size[0] * self.window_size[1],
-1,
) # Wh*Ww,Wh*Ww,nH
relative_position_bias = relative_position_bias.permute(
2, 0, 1
).contiguous() # nH, Wh*Ww, Wh*Ww
attn = attn + relative_position_bias.unsqueeze(0)
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)
x = (attn @ v).transpose(1, 2).reshape(B_, N, C)
x = self.proj(x)
x = self.proj_drop(x)
return x补丁合并层
块合并方法用于下采样。它用于降低特征图的空间维度，类似于传统卷积神经网络 (CNN) 中的池化操作。它通过逐步增加感受野并降低空间分辨率来帮助构建分层特征表示。
python
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().__init__()
self.input_resolution = input_resolution
self.dim = dim
self.reduction = nn.Linear(4 * dim, 2 * dim, bias=False)
self.norm = norm_layer(4 * dim)
def forward(self, x):
"""
x: B, H*W, C
"""
H, W = self.input_resolution
B, L, C = x.shape
assert 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 C
x1 = x[:, 1::2, 0::2, :] # B H/2 W/2 C
x2 = x[:, 0::2, 1::2, :] # B H/2 W/2 C
x3 = x[:, 1::2, 1::2, :] # B H/2 W/2 C
x = torch.cat([x0, x1, x2, x3], -1) # B H/2 W/2 4*C
x = x.view(B, -1, 4 * C) # B H/2*W/2 4*C
x = self.norm(x)
x = self.reduction(x)
return x五、使用预训练的Swin模型进行分类
ini
from datasets import load_dataset
from transformers import AutoImageProcessor, SwinForImageClassification
import torch
model = SwinForImageClassification.from_pretrained(
"microsoft/swin-tiny-patch4-window7-224"
)
image_processor = AutoImageProcessor.from_pretrained(
"microsoft/swin-tiny-patch4-window7-224"
)
dataset = load_dataset("huggingface/cats-image")
image = dataset["test"]["image"][0]
inputs = image_processor(image, return_tensors="pt")
with torch.no_grad():
logits = model(**inputs).logits
predicted_label_id = logits.argmax(-1).item()
predicted_label_text = model.config.id2label[predicted_label_id]
print(predicted_label_text)六、优劣分析与最佳实践
- 位置编码：启用绝对位置嵌入（ape=True）提升小目标检测
- 深度衰减：使用drop_path_rate防止深层过拟合
- 硬件加速：开启fused_window_process优化窗口移位计算
结语
Swin Transformer通过"局部计算，全局通信" 的设计哲学，成功打通视觉Transformer的高效之路。随着SwinV2支持1.5K分辨率图像和30亿参数模型，其已成为工业级视觉任务的基石架构。当卷积遭遇注意力，当局部邂逅全局------Swin用移位窗口打开了视觉认知的新视界。