【ICCV2021】Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions

Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions

Interpretation: Vernacular Pyramid Vision Transformer - Zhihu (zhihu.com)

PVT: Pyramid visual transformer for intensive task backbone - Zhihu (zhihu.com)

introduce

Summary: PVT introduces the pyramid structure into Transformer so that it can be seamlessly connected to various downstream tasks. Simply adjust Multi-Head Attention and propose spatial reduction attention.

This article imitates the common pyramid structure in CNNs, improves the original Transformer, divides multiple stages, halve the length and width of each stage, increases the channel dimension, and then superimposes multiple stages. It can be applied to tasks such as classification, detection, and segmentation.

method

PVT network

The network is divided into 4 stages. The input of each stage is a 3D feature map. At the beginning of each stage, the input image or feature is first tokenized like ViT, that is, patch embedding is performed. The patch size is 2x2 (the first stage is 4*4), which means the final feature map size of the stage. Halved, the number of tokens is reduced by 4 times. PVT has a total of 4 stages, and the feature maps obtained by the 4 stages are 1/4, 1/8, 1/16 and 1/32 respectively compared to the original image size. The number of tokens in different stages is different. Each stage uses different position embeddings. Each stage is added with its own position embedding after the patch embed. When the input image size changes, the position embeddings can also be adapted through interpolation.

Network structure

Details are shown in the table below. There are 4 variants.

SRA

To further reduce the amount of calculation, conventional multi-head attention (MHA) is replaced with spatial-reduction attention (SRA). The core of SRA is to reduce the number of key and value pairs in the attention layer. When conventional MHA is calculated in the attention layer, the number of key and value pairs is the length of the sequence, but SRA reduces it to the original $1/R^2$ .

key code

Attention

# https://github.com/whai362/PVT/blob/v2/segmentation/pvt.py

class Attention(nn.Module):
    def __init__(self, dim, num_heads=8, qkv_bias=False, qk_scale=None, attn_drop=0., proj_drop=0., sr_ratio=1):
        super().__init__()
        assert dim % num_heads == 0, f"dim {dim} should be divided by num_heads {num_heads}."

        self.dim = dim
        self.num_heads = num_heads
        head_dim = dim // num_heads
        self.scale = qk_scale or head_dim ** -0.5

        self.q = nn.Linear(dim, dim, bias=qkv_bias)
        self.kv = nn.Linear(dim, dim * 2, bias=qkv_bias)
        self.attn_drop = nn.Dropout(attn_drop)
        self.proj = nn.Linear(dim, dim)
        self.proj_drop = nn.Dropout(proj_drop)

        self.sr_ratio = sr_ratio
        # 实现上这里等价于一个卷积层
        if sr_ratio > 1:
            self.sr = nn.Conv2d(dim, dim, kernel_size=sr_ratio, stride=sr_ratio)
            self.norm = nn.LayerNorm(dim)

    def forward(self, x, H, W):
        B, N, C = x.shape
        q = self.q(x).reshape(B, N, self.num_heads, C // self.num_heads).permute(0, 2, 1, 3)

        if self.sr_ratio > 1:
            x_ = x.permute(0, 2, 1).reshape(B, C, H, W)
            x_ = self.sr(x_).reshape(B, C, -1).permute(0, 2, 1) # 这里x_.shape = (B, N/R^2, C)
            x_ = self.norm(x_)
            kv = self.kv(x_).reshape(B, -1, 2, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)
        else:
            kv = self.kv(x).reshape(B, -1, 2, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)
        k, v = kv[0], kv[1]

        attn = (q @ k.transpose(-2, -1)) * self.scale
        attn = attn.softmax(dim=-1)
        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

PVT

# https://github.com/whai362/PVT/blob/v2/segmentation/pvt.py


import torch
import torch.nn as nn
import torch.nn.functional as F
from functools import partial

from timm.models.layers import DropPath, to_2tuple, trunc_normal_
from timm.models.registry import register_model
from timm.models.vision_transformer import _cfg
from mmseg.models.builder import BACKBONES
from mmseg.utils import get_root_logger
from mmcv.runner import load_checkpoint


class Mlp(nn.Module):
    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.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


class Attention(nn.Module):
    def __init__(self, dim, num_heads=8, qkv_bias=False, qk_scale=None, attn_drop=0., proj_drop=0., sr_ratio=1):
        super().__init__()
        assert dim % num_heads == 0, f"dim {dim} should be divided by num_heads {num_heads}."

        self.dim = dim
        self.num_heads = num_heads
        head_dim = dim // num_heads
        self.scale = qk_scale or head_dim ** -0.5

        self.q = nn.Linear(dim, dim, bias=qkv_bias)
        self.kv = nn.Linear(dim, dim * 2, bias=qkv_bias)
        self.attn_drop = nn.Dropout(attn_drop)
        self.proj = nn.Linear(dim, dim)
        self.proj_drop = nn.Dropout(proj_drop)

        self.sr_ratio = sr_ratio
        if sr_ratio > 1:
            self.sr = nn.Conv2d(dim, dim, kernel_size=sr_ratio, stride=sr_ratio)
            self.norm = nn.LayerNorm(dim)

    def forward(self, x, H, W):
        B, N, C = x.shape
        q = self.q(x).reshape(B, N, self.num_heads, C // self.num_heads).permute(0, 2, 1, 3)

        if self.sr_ratio > 1:
            x_ = x.permute(0, 2, 1).reshape(B, C, H, W)
            x_ = self.sr(x_).reshape(B, C, -1).permute(0, 2, 1)
            x_ = self.norm(x_)
            kv = self.kv(x_).reshape(B, -1, 2, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)
        else:
            kv = self.kv(x).reshape(B, -1, 2, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)
        k, v = kv[0], kv[1]

        attn = (q @ k.transpose(-2, -1)) * self.scale
        attn = attn.softmax(dim=-1)
        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


class Block(nn.Module):

    def __init__(self, dim, num_heads, mlp_ratio=4., qkv_bias=False, qk_scale=None, drop=0., attn_drop=0.,
                 drop_path=0., act_layer=nn.GELU, norm_layer=nn.LayerNorm, sr_ratio=1):
        super().__init__()
        self.norm1 = norm_layer(dim)
        self.attn = Attention(
            dim,
            num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale,
            attn_drop=attn_drop, proj_drop=drop, sr_ratio=sr_ratio)
        # NOTE: drop path for stochastic depth, we shall see if this is better than dropout here
        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, H, W):
        x = x + self.drop_path(self.attn(self.norm1(x), H, W))
        x = x + self.drop_path(self.mlp(self.norm2(x)))

        return x


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

    def __init__(self, img_size=224, patch_size=16, in_chans=3, embed_dim=768):
        super().__init__()
        img_size = to_2tuple(img_size)
        patch_size = to_2tuple(patch_size)

        self.img_size = img_size
        self.patch_size = patch_size
        assert img_size[0] % patch_size[0] == 0 and img_size[1] % patch_size[1] == 0, \
            f"img_size {img_size} should be divided by patch_size {patch_size}."
        self.H, self.W = img_size[0] // patch_size[0], img_size[1] // patch_size[1]
        self.num_patches = self.H * self.W
        self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_size, stride=patch_size)
        self.norm = nn.LayerNorm(embed_dim)

    def forward(self, x):
        B, C, H, W = x.shape

        x = self.proj(x).flatten(2).transpose(1, 2)
        x = self.norm(x)
        H, W = H // self.patch_size[0], W // self.patch_size[1]

        return x, (H, W)


class PyramidVisionTransformer(nn.Module):
    def __init__(self, img_size=224, patch_size=16, in_chans=3, num_classes=1000, embed_dims=[64, 128, 256, 512],
                 num_heads=[1, 2, 4, 8], mlp_ratios=[4, 4, 4, 4], qkv_bias=False, qk_scale=None, drop_rate=0.,
                 attn_drop_rate=0., drop_path_rate=0., norm_layer=nn.LayerNorm, depths=[3, 4, 6, 3],
                 sr_ratios=[8, 4, 2, 1], num_stages=4, F4=False):
        super().__init__()
        self.num_classes = num_classes
        self.depths = depths
        self.F4 = F4
        self.num_stages = num_stages

        dpr = [x.item() for x in torch.linspace(0, drop_path_rate, sum(depths))]  # stochastic depth decay rule
        cur = 0

        for i in range(num_stages):
            patch_embed = PatchEmbed(img_size=img_size if i == 0 else img_size // (2 ** (i + 1)),
                                     patch_size=patch_size if i == 0 else 2,
                                     in_chans=in_chans if i == 0 else embed_dims[i - 1],
                                     embed_dim=embed_dims[i])
            num_patches = patch_embed.num_patches if i != num_stages - 1 else patch_embed.num_patches + 1
            pos_embed = nn.Parameter(torch.zeros(1, num_patches, embed_dims[i]))
            pos_drop = nn.Dropout(p=drop_rate)

            block = nn.ModuleList([Block(
                dim=embed_dims[i], num_heads=num_heads[i], mlp_ratio=mlp_ratios[i], qkv_bias=qkv_bias,
                qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dpr[cur + j],
                norm_layer=norm_layer, sr_ratio=sr_ratios[i])
                for j in range(depths[i])])
            cur += depths[i]

            setattr(self, f"patch_embed{i + 1}", patch_embed)
            setattr(self, f"pos_embed{i + 1}", pos_embed)
            setattr(self, f"pos_drop{i + 1}", pos_drop)
            setattr(self, f"block{i + 1}", block)

            trunc_normal_(pos_embed, std=.02)

        # init weights
        self.apply(self._init_weights)

    def init_weights(self, pretrained=None):
        if isinstance(pretrained, str):
            logger = get_root_logger()
            load_checkpoint(self, pretrained, map_location='cpu', strict=False, logger=logger)

    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)

    def _get_pos_embed(self, pos_embed, patch_embed, H, W):
        if H * W == self.patch_embed1.num_patches:
            return pos_embed
        else:
            return F.interpolate(
                pos_embed.reshape(1, patch_embed.H, patch_embed.W, -1).permute(0, 3, 1, 2),
                size=(H, W), mode="bilinear").reshape(1, -1, H * W).permute(0, 2, 1)

    def forward_features(self, x):
        outs = []

        B = x.shape[0]

        for i in range(self.num_stages):
            patch_embed = getattr(self, f"patch_embed{i + 1}")
            pos_embed = getattr(self, f"pos_embed{i + 1}")
            pos_drop = getattr(self, f"pos_drop{i + 1}")
            block = getattr(self, f"block{i + 1}")
            x, (H, W) = patch_embed(x)
            if i == self.num_stages - 1:
                pos_embed = self._get_pos_embed(pos_embed[:, 1:], patch_embed, H, W)
            else:
                pos_embed = self._get_pos_embed(pos_embed, patch_embed, H, W)

            x = pos_drop(x + pos_embed)
            for blk in block:
                x = blk(x, H, W)
            x = x.reshape(B, H, W, -1).permute(0, 3, 1, 2).contiguous()
            outs.append(x)

        return outs

    def forward(self, x):
        x = self.forward_features(x)

        if self.F4:
            x = x[3:4]

        return x


def _conv_filter(state_dict, patch_size=16):
    """ convert patch embedding weight from manual patchify + linear proj to conv"""
    out_dict = {}
    for k, v in state_dict.items():
        if 'patch_embed.proj.weight' in k:
            v = v.reshape((v.shape[0], 3, patch_size, patch_size))
        out_dict[k] = v

    return out_dict

4 variations

# https://github.com/whai362/PVT/blob/v2/segmentation/pvt.py


@BACKBONES.register_module()
class pvt_tiny(PyramidVisionTransformer):
    def __init__(self, **kwargs):
        super(pvt_tiny, self).__init__(
            patch_size=4, embed_dims=[64, 128, 320, 512], num_heads=[1, 2, 5, 8], mlp_ratios=[8, 8, 4, 4],
            qkv_bias=True, norm_layer=partial(nn.LayerNorm, eps=1e-6), depths=[2, 2, 2, 2],
            sr_ratios=[8, 4, 2, 1], drop_rate=0.0, drop_path_rate=0.1)


@BACKBONES.register_module()
class pvt_small(PyramidVisionTransformer):
    def __init__(self, **kwargs):
        super(pvt_small, self).__init__(
            patch_size=4, embed_dims=[64, 128, 320, 512], num_heads=[1, 2, 5, 8], mlp_ratios=[8, 8, 4, 4],
            qkv_bias=True, norm_layer=partial(nn.LayerNorm, eps=1e-6), depths=[3, 4, 6, 3],
            sr_ratios=[8, 4, 2, 1], drop_rate=0.0, drop_path_rate=0.1)


@BACKBONES.register_module()
class pvt_medium(PyramidVisionTransformer):
    def __init__(self, **kwargs):
        super(pvt_medium, self).__init__(
            patch_size=4, embed_dims=[64, 128, 320, 512], num_heads=[1, 2, 5, 8], mlp_ratios=[8, 8, 4, 4],
            qkv_bias=True, norm_layer=partial(nn.LayerNorm, eps=1e-6), depths=[3, 4, 18, 3],
            sr_ratios=[8, 4, 2, 1], drop_rate=0.0, drop_path_rate=0.1)


@BACKBONES.register_module()
class pvt_large(PyramidVisionTransformer):
    def __init__(self, **kwargs):
        super(pvt_large, self).__init__(
            patch_size=4, embed_dims=[64, 128, 320, 512], num_heads=[1, 2, 5, 8], mlp_ratios=[8, 8, 4, 4],
            qkv_bias=True, norm_layer=partial(nn.LayerNorm, eps=1e-6), depths=[3, 8, 27, 3],
            sr_ratios=[8, 4, 2, 1], drop_rate=0.0, drop_path_rate=0.1)