第十五周周报

处理代码的前向计算问题，因为使用的语义图卷积和transformer是用于人体姿势估计的，其中的维度大小和手姿势估计不匹配，要进行修改。主要是以下内容：

1、更改transformer的框架。将时间流去掉，所用到的原代码是时间—空间的transformer的人体姿态预测，删除了与时间维度相关的代码，修改后的代码如下：

## Our PoseFormer model was revised from https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/vision_transformer.py

import math
import logging
from functools import partial
from collections import OrderedDict
from einops import rearrange, repeat

import torch
import torch.nn as nn
import torch.nn.functional as F

from timm.data import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD
from timm.models.helpers import load_pretrained
from timm.models.layers import DropPath, to_2tuple, trunc_normal_
from timm.models.registry import register_model


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.):
        super().__init__()
        self.num_heads = num_heads
        head_dim = dim // num_heads
        # NOTE scale factor was wrong in my original version, can set manually to be compat with prev weights
        self.scale = qk_scale or head_dim ** -0.5

        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)

    def forward(self, x):
        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)

        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):
        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)
        # 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):
        x = x + self.drop_path(self.attn(self.norm1(x)))
        x = x + self.drop_path(self.mlp(self.norm2(x)))
        return x


class PoseTransformer(nn.Module):
    def __init__(
        self, num_joints=17, in_chans=2, embed_dim_ratio=32, depth=4,
        num_heads=8, mlp_ratio=2., qkv_bias=True, qk_scale=None,
        drop_rate=0., attn_drop_rate=0., drop_path_rate=0.2, norm_layer=None
    ):
        super().__init__()

        norm_layer = norm_layer or partial(nn.LayerNorm, eps=1e-6)
        embed_dim = embed_dim_ratio * num_joints
        # out_dim = num_joints * 4
        out_dim = embed_dim_ratio * 4

        self.Spatial_patch_to_embedding = nn.Linear(in_chans, embed_dim)
        self.Spatial_pos_embed = nn.Parameter(torch.zeros(1, num_joints, embed_dim))
        self.pos_drop = nn.Dropout(p=drop_rate)

        dpr = [x.item() for x in torch.linspace(0, drop_path_rate, depth)]

        self.Spatial_blocks = nn.ModuleList([
            Block(
                dim=embed_dim, num_heads=num_heads, mlp_ratio=mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale,
                drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dpr[i], norm_layer=norm_layer
            )
            for i in range(depth)
        ])

        self.Spatial_norm = norm_layer(embed_dim)

        self.head = nn.Sequential(
            nn.LayerNorm(embed_dim),
            nn.Linear(embed_dim, out_dim),
        )

    def forward(self, x):
        x = x.permute(0, 3, 1, 2)
        b, _, _, p = x.shape
        x = rearrange(x, 'b c f p -> (b f) p c')

        x = self.Spatial_patch_to_embedding(x)
        x += self.Spatial_pos_embed
        x = self.pos_drop(x)

        for blk in self.Spatial_blocks:
            x = blk(x)

        x = self.Spatial_norm(x)
        x = rearrange(x, '(b f) p c -> b f p c', f=1)
        x = self.head(x)

        x = x.view(b, 1, p, -1)

        return x

2、修改前向代码，由于前向代码在计算时会出现维度不匹配的情况，需要转换张量的维度，这里需要消耗很多时间！可以选择边调试forward过程边修改结果，介绍debug前向过程的方法:

(1条消息) 如何调试类中的forward函数？_forward函数怎么调用_Faster--YOLO的博客-CSDN博客

可以参考上面的做法

3、待解决的问题：

①前向过程的计算，如张量维度进行矩阵乘法时维度不匹配的问题，需要仔细比较传进去的参数，看维度大小的变化

②人体姿势估计的代码用于手估计还是有很大的差别，所以不太好修改，另外，这两个模块的代码都是用于人体的源代码，只能根据源代码进行修改，用于手的参考论文中还没有释放这个代码，所以花费大量的时间。