SwinIR代码理解笔记

jackhalifa

已于 2023-12-20 21:39:16 修改

阅读量1.2k

点赞数 19

文章标签：笔记人工智能 transformer

于 2023-12-20 19:38:59 首次发布

本文链接：https://blog.csdn.net/qq_39272190/article/details/135100075

版权

最近研究了SwinIR的原理和代码，把自己的理解记录下来，用作分享。

如果有理解的不准确的地方，欢迎指正交流。

论文地址：https://arxiv.org/abs/2108.10257

源代码地址：https://github.com/JingyunLiang/SwinIR

一、模型简介

SwinIR的的网络由三部分组成：浅层特征提取，深层特征提取以及图像重建。

浅层特征提取：利用一个 $3\times 3$ 的卷积来提取浅层特征

深层特征提取：利用 $K$ 个残差Swin Transformer块和一个卷积模块实现深层特征提取

图像重建：将浅层和深层特征通过跳连接输入图像重建模块，浅层特征主要包含低频信息，而深层特征侧重于恢复丢失的高频。通过跳连接，SwinIR可以将低频信息直接传递给重构模块，帮助深度特征提取模块专注于高频信息，稳定训练。对于超分辨率、去噪等任务，图像重建模块的实现方法不同，具体请查阅具体代码。

二、关键代码理解

1.Patch Embedding & Patch Unembedding

对于SwinIR来说，对于Patch的操作和基础的Swin Transformer还是有一定的区别的。

在Swin Transformer中是通过一个卷积将输入图片的维度扩展，再分成Patch输入不同阶段（stage）的Transformer Block中。

而对于SwinIR，由于浅层特征已经提取完毕，这里的Patch Embbedding只是将特征的维度转换，由[B, C, H, W]转换为[B, Ph*Pw, C]。而Patch Unembedding操作是将三维的特征[B, Ph*Pw, C]重新转换为四维的[B, C, H, W]。

这两个操作是成对出现的，每次进入RTSB模块前都会引入Patch Embbedding操作，而在RTSB中最后一个Transformer层之后都会再进行Patch Unembedding之后再输入RTSB模块最后的卷积层。

代码如下：

class PatchEmbed(nn.Module):
    r""" Image to Patch Embedding

    Args:
        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().__init__()
        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_size
        self.patch_size = patch_size
        self.patches_resolution = patches_resolution
        self.num_patches = patches_resolution[0] * patches_resolution[1]
        print(self.num_patches)
        self.in_chans = in_chans
        self.embed_dim = embed_dim
        if norm_layer is not None:
            self.norm = norm_layer(embed_dim)
        else:
            self.norm = None

    def forward(self, x):
        x = x.flatten(2).transpose(1, 2)  # B Ph*Pw C
        print(x.shape)
        if self.norm is not None:
            x = self.norm(x)
        return x

    def flops(self):
        flops = 0
        H, W = self.img_size
        if self.norm is not None:
            flops += H * W * self.embed_dim
        return flops


class PatchUnEmbed(nn.Module):
    r""" Image to Patch Unembedding

    Args:
        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().__init__()
        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_size
        self.patch_size = patch_size
        self.patches_resolution = patches_resolution
        self.num_patches = patches_resolution[0] * patches_resolution[1]

        self.in_chans = in_chans
        self.embed_dim = embed_dim

    def forward(self, x, x_size):
        B, HW, C = x.shape
        x = x.transpose(1, 2).view(B, self.embed_dim, x_size[0], x_size[1])  # B Ph*Pw C
        print(x.shape)
        return x

    def flops(self):
        flops = 0
        return flops

另外值得注意的是，在SwinIR具体实现过程中，所有的patch_size设定的均为1。

也就是说SwinIR的操作都是在像素层级（pixel-wise）进行的，而没有先划分patch块再进行Transformer的自注意力计算。

2.Swin Transformer Layer

这部分代码结构与Swin Transformer原文中实现方法与原理一致。

代码如下：

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
    """

    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().__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. 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:
            attn_mask = self.calculate_mask(self.input_resolution)
        else:
            attn_mask = None

        self.register_buffer("attn_mask", attn_mask)

    def calculate_mask(self, x_size):
        # calculate attention mask for SW-MSA
        H, W = x_size
        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))

        return attn_mask

    def forward(self, x, x_size):
        H, W = x_size
        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:
            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.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 (to be compatible for testing on images whose shapes are the multiple of window size
        if self.input_resolution == x_size:
            attn_windows = self.attn(x_windows, mask=self.attn_mask)  # nW*B, window_size*window_size, C
        else:
            attn_windows = self.attn(x_windows, mask=self.calculate_mask(x_size).to(x.device))

        # merge windows
        attn_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 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)))

        return x

    def 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 = 0
        H, W = self.input_resolution
        # norm1
        flops += self.dim * H * W
        # W-MSA/SW-MSA
        nW = H * W / self.window_size / self.window_size
        flops += nW * self.attn.flops(self.window_size * self.window_size)
        # mlp
        flops += 2 * H * W * self.dim * self.dim * self.mlp_ratio
        # norm2
        flops += self.dim * H * W
        return flops

在这段代码当中核心的设计是移动窗口注意力掩码的设计，注意其中有应用到广播机制：

    def calculate_mask(self, x_size):
        # calculate attention mask for SW-MSA
        H, W = x_size
        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)
        # 广播机制 [nW,1,window_size*window_size]-[nW,window_size*window_size,1]
        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))

        return attn_mask

构建完掩码之后，要将掩码进行没有重叠的窗口切分（Window Partition）。

如下代码介绍了如何进行窗口切分（Window Partition）和将窗口还原成特征图的过程（Window Reserve）：

def window_partition(x, window_size):
    """
    Args:
        x: (B, H, W, C)
        window_size (int): window size

    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)
    windows = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(-1, window_size, window_size, C)
    return windows


def window_reverse(windows, window_size, H, W):
    """
    Args:
        windows: (num_windows*B, window_size, window_size, C)
        window_size (int): Window size
        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))
    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

另外需要注意的是在偶数个swin transformer层中，需要对特征图左和上的特征向右和向下平移，这一操作是通过torch的roll函数实现的

if self.shift_size > 0:
            x = torch.roll(shifted_x, shifts=(self.shift_size, self.shift_size), dims=(1, 2))

3. Window Attention

窗口自注意力的计算，此部分与普通的vision transformer的区别是，使用了相对位置的计算。

代码如下：

# 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)