EasyMocap/myeasymocap/backbone/vitpose/vit_moe.py
2023-07-10 22:10:41 +08:00

608 lines
22 KiB
Python

# Copyright (c) OpenMMLab. All rights reserved.
import os
import numpy as np
import torch
from functools import partial
import torch.nn as nn
import torch.nn.functional as F
import torch.utils.checkpoint as checkpoint
from .layers import drop_path, to_2tuple, trunc_normal_
def drop_path(x, drop_prob: float = 0., training: bool = False, scale_by_keep: bool = True):
"""Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks).
This is the same as the DropConnect impl I created for EfficientNet, etc networks, however,
the original name is misleading as 'Drop Connect' is a different form of dropout in a separate paper...
See discussion: https://github.com/tensorflow/tpu/issues/494#issuecomment-532968956 ... I've opted for
changing the layer and argument names to 'drop path' rather than mix DropConnect as a layer name and use
'survival rate' as the argument.
"""
if drop_prob == 0. or not training:
return x
keep_prob = 1 - drop_prob
shape = (x.shape[0],) + (1,) * (x.ndim - 1) # work with diff dim tensors, not just 2D ConvNets
random_tensor = x.new_empty(shape).bernoulli_(keep_prob)
if keep_prob > 0.0 and scale_by_keep:
random_tensor.div_(keep_prob)
return x * random_tensor
def get_abs_pos(abs_pos, h, w, ori_h, ori_w, has_cls_token=True):
"""
Calculate absolute positional embeddings. If needed, resize embeddings and remove cls_token
dimension for the original embeddings.
Args:
abs_pos (Tensor): absolute positional embeddings with (1, num_position, C).
has_cls_token (bool): If true, has 1 embedding in abs_pos for cls token.
hw (Tuple): size of input image tokens.
Returns:
Absolute positional embeddings after processing with shape (1, H, W, C)
"""
cls_token = None
B, L, C = abs_pos.shape
if has_cls_token:
cls_token = abs_pos[:, 0:1]
abs_pos = abs_pos[:, 1:]
if ori_h != h or ori_w != w:
new_abs_pos = F.interpolate(
abs_pos.reshape(1, ori_h, ori_w, -1).permute(0, 3, 1, 2),
size=(h, w),
mode="bicubic",
align_corners=False,
).permute(0, 2, 3, 1).reshape(B, -1, C)
else:
new_abs_pos = abs_pos
if cls_token is not None:
new_abs_pos = torch.cat([cls_token, new_abs_pos], dim=1)
return new_abs_pos
class DropPath(nn.Module):
"""Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks).
"""
def __init__(self, drop_prob=None):
super(DropPath, self).__init__()
self.drop_prob = drop_prob
def forward(self, x):
return drop_path(x, self.drop_prob, self.training)
def extra_repr(self):
return 'p={}'.format(self.drop_prob)
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.fc2(x)
x = self.drop(x)
return x
class MoEMlp(nn.Module):
def __init__(self, num_expert=1, in_features=1024, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0., part_features=256):
super().__init__()
out_features = out_features or in_features
hidden_features = hidden_features or in_features
self.part_features = part_features
self.fc1 = nn.Linear(in_features, hidden_features)
self.act = act_layer()
self.fc2 = nn.Linear(hidden_features, out_features - part_features)
self.drop = nn.Dropout(drop)
self.num_expert = num_expert
experts = []
for i in range(num_expert):
experts.append(
nn.Linear(hidden_features, part_features)
)
self.experts = nn.ModuleList(experts)
def forward(self, x, indices):
expert_x = torch.zeros_like(x[:, :, -self.part_features:], device=x.device, dtype=x.dtype)
x = self.fc1(x)
x = self.act(x)
shared_x = self.fc2(x)
indices = indices.view(-1, 1, 1)
# to support ddp training
for i in range(self.num_expert):
selectedIndex = (indices == i)
current_x = self.experts[i](x) * selectedIndex
expert_x = expert_x + current_x
x = torch.cat([shared_x, expert_x], dim=-1)
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., attn_head_dim=None,):
super().__init__()
self.num_heads = num_heads
head_dim = dim // num_heads
self.dim = dim
if attn_head_dim is not None:
head_dim = attn_head_dim
all_head_dim = head_dim * self.num_heads
self.scale = qk_scale or head_dim ** -0.5
self.qkv = nn.Linear(dim, all_head_dim * 3, bias=qkv_bias)
self.attn_drop = nn.Dropout(attn_drop)
self.proj = nn.Linear(all_head_dim, dim)
self.proj_drop = nn.Dropout(proj_drop)
def forward(self, x):
B, N, C = x.shape
qkv = self.qkv(x)
qkv = qkv.reshape(B, N, 3, self.num_heads, -1).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))
attn = attn.softmax(dim=-1)
attn = self.attn_drop(attn)
x = (attn @ v).transpose(1, 2).reshape(B, N, -1)
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, attn_head_dim=None, num_expert=1, part_features=None
):
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, attn_head_dim=attn_head_dim
)
# 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 = MoEMlp(num_expert=num_expert, in_features=dim, hidden_features=mlp_hidden_dim,
act_layer=act_layer, drop=drop, part_features=part_features)
def forward(self, x, indices=None):
x = x + self.drop_path(self.attn(self.norm1(x)))
x = x + self.drop_path(self.mlp(self.norm2(x), indices))
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, ratio=1):
super().__init__()
img_size = to_2tuple(img_size)
patch_size = to_2tuple(patch_size)
num_patches = (img_size[1] // patch_size[1]) * (img_size[0] // patch_size[0]) * (ratio ** 2)
self.patch_shape = (int(img_size[0] // patch_size[0] * ratio), int(img_size[1] // patch_size[1] * ratio))
self.origin_patch_shape = (int(img_size[0] // patch_size[0]), int(img_size[1] // patch_size[1]))
self.img_size = img_size
self.patch_size = patch_size
self.num_patches = num_patches
self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_size, stride=(patch_size[0] // ratio), padding=4 + 2 * (ratio//2-1))
def forward(self, x, **kwargs):
B, C, H, W = x.shape
x = self.proj(x)
Hp, Wp = x.shape[2], x.shape[3]
x = x.flatten(2).transpose(1, 2)
return x, (Hp, Wp)
class HybridEmbed(nn.Module):
""" CNN Feature Map Embedding
Extract feature map from CNN, flatten, project to embedding dim.
"""
def __init__(self, backbone, img_size=224, feature_size=None, in_chans=3, embed_dim=768):
super().__init__()
assert isinstance(backbone, nn.Module)
img_size = to_2tuple(img_size)
self.img_size = img_size
self.backbone = backbone
if feature_size is None:
with torch.no_grad():
training = backbone.training
if training:
backbone.eval()
o = self.backbone(torch.zeros(1, in_chans, img_size[0], img_size[1]))[-1]
feature_size = o.shape[-2:]
feature_dim = o.shape[1]
backbone.train(training)
else:
feature_size = to_2tuple(feature_size)
feature_dim = self.backbone.feature_info.channels()[-1]
self.num_patches = feature_size[0] * feature_size[1]
self.proj = nn.Linear(feature_dim, embed_dim)
def forward(self, x):
x = self.backbone(x)[-1]
x = x.flatten(2).transpose(1, 2)
x = self.proj(x)
return x
class ViTMoE(nn.Module):
def __init__(self,
img_size=224, patch_size=16, in_chans=3, num_classes=80, embed_dim=768, depth=12,
num_heads=12, mlp_ratio=4., qkv_bias=False, qk_scale=None, drop_rate=0., attn_drop_rate=0.,
drop_path_rate=0., hybrid_backbone=None, norm_layer=None, use_checkpoint=False,
frozen_stages=-1, ratio=1, last_norm=True,
patch_padding='pad', freeze_attn=False, freeze_ffn=False,
num_expert=1, part_features=None
):
# Protect mutable default arguments
super(ViTMoE, self).__init__()
norm_layer = norm_layer or partial(nn.LayerNorm, eps=1e-6)
self.num_classes = num_classes
self.num_features = self.embed_dim = embed_dim # num_features for consistency with other models
self.frozen_stages = frozen_stages
self.use_checkpoint = use_checkpoint
self.patch_padding = patch_padding
self.freeze_attn = freeze_attn
self.freeze_ffn = freeze_ffn
self.depth = depth
if hybrid_backbone is not None:
self.patch_embed = HybridEmbed(
hybrid_backbone, img_size=img_size, in_chans=in_chans, embed_dim=embed_dim)
else:
self.patch_embed = PatchEmbed(
img_size=img_size, patch_size=patch_size, in_chans=in_chans, embed_dim=embed_dim, ratio=ratio)
num_patches = self.patch_embed.num_patches
self.part_features = part_features
self.pos_embed = nn.Parameter(torch.zeros(1, num_patches + 1, embed_dim))
dpr = [x.item() for x in torch.linspace(0, drop_path_rate, depth)] # stochastic depth decay rule
self.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,
num_expert=num_expert, part_features=part_features
)
for i in range(depth)])
self.last_norm = norm_layer(embed_dim) if last_norm else nn.Identity()
if self.pos_embed is not None:
trunc_normal_(self.pos_embed, std=.02)
self._freeze_stages()
def _freeze_stages(self):
"""Freeze parameters."""
if self.frozen_stages >= 0:
self.patch_embed.eval()
for param in self.patch_embed.parameters():
param.requires_grad = False
for i in range(1, self.frozen_stages + 1):
m = self.blocks[i]
m.eval()
for param in m.parameters():
param.requires_grad = False
if self.freeze_attn:
for i in range(0, self.depth):
m = self.blocks[i]
m.attn.eval()
m.norm1.eval()
for param in m.attn.parameters():
param.requires_grad = False
for param in m.norm1.parameters():
param.requires_grad = False
if self.freeze_ffn:
self.pos_embed.requires_grad = False
self.patch_embed.eval()
for param in self.patch_embed.parameters():
param.requires_grad = False
for i in range(0, self.depth):
m = self.blocks[i]
m.mlp.eval()
m.norm2.eval()
for param in m.mlp.parameters():
param.requires_grad = False
for param in m.norm2.parameters():
param.requires_grad = False
def init_weights(self, pretrained=None):
"""Initialize the weights in backbone.
Args:
pretrained (str, optional): Path to pre-trained weights.
Defaults to None.
"""
super().init_weights(pretrained, patch_padding=self.patch_padding, part_features=self.part_features)
if pretrained is None:
def _init_weights(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)
self.apply(_init_weights)
def get_num_layers(self):
return len(self.blocks)
@torch.jit.ignore
def no_weight_decay(self):
return {'pos_embed', 'cls_token'}
def forward_features(self, x, dataset_source=None):
B, C, H, W = x.shape
x, (Hp, Wp) = self.patch_embed(x)
if self.pos_embed is not None:
# fit for multiple GPU training
# since the first element for pos embed (sin-cos manner) is zero, it will cause no difference
x = x + self.pos_embed[:, 1:] + self.pos_embed[:, :1]
for blk in self.blocks:
if self.use_checkpoint:
x = checkpoint.checkpoint(blk, x, dataset_source)
else:
x = blk(x, dataset_source)
x = self.last_norm(x)
xp = x.permute(0, 2, 1).reshape(B, -1, Hp, Wp).contiguous()
return xp
def forward(self, x, dataset_source=None):
x = self.forward_features(x, dataset_source)
return x
def train(self, mode=True):
"""Convert the model into training mode."""
super().train(mode)
self._freeze_stages()
class Head(nn.Module):
def __init__(self, in_channels,
out_channels,
num_deconv_layers=3,
num_deconv_filters=(256, 256, 256),
num_deconv_kernels=(4, 4, 4),):
super().__init__()
self.in_channels = in_channels
self.deconv_layers = self._make_deconv_layer(num_deconv_layers, num_deconv_filters, num_deconv_kernels)
self.final_layer = nn.Conv2d(in_channels=num_deconv_filters[-1], out_channels=out_channels,
kernel_size=1, stride=1, padding=0)
def _make_deconv_layer(self, num_layers, num_filters, num_kernels):
layers = []
for i in range(num_layers):
kernel, padding, output_padding = \
self._get_deconv_cfg(num_kernels[i])
planes = num_filters[i]
layers.append(
nn.ConvTranspose2d(
in_channels=self.in_channels,
out_channels=planes,
kernel_size=kernel,
stride=2,
padding=padding,
output_padding=output_padding,
bias=False))
layers.append(nn.BatchNorm2d(planes))
layers.append(nn.ReLU(inplace=True))
self.in_channels = planes
return nn.Sequential(*layers)
@staticmethod
def _get_deconv_cfg(deconv_kernel):
"""Get configurations for deconv layers."""
if deconv_kernel == 4:
padding = 1
output_padding = 0
elif deconv_kernel == 3:
padding = 1
output_padding = 1
elif deconv_kernel == 2:
padding = 0
output_padding = 0
else:
raise ValueError(f'Not supported num_kernels ({deconv_kernel}).')
return deconv_kernel, padding, output_padding
def forward(self, x):
"""Forward function."""
x = self.deconv_layers(x)
x = self.final_layer(x)
return x
class ComposeVit(nn.Module):
def __init__(self):
super().__init__()
cfg_backbone = dict(
img_size=(256, 192),
patch_size=16,
embed_dim=768,
depth=12,
num_heads=12,
ratio=1,
use_checkpoint=False,
mlp_ratio=4,
qkv_bias=True,
drop_path_rate=0.3,
num_expert=6,
part_features=192
)
cfg_head = dict(
in_channels=768,
out_channels=17,
num_deconv_layers=2,
num_deconv_filters=(256, 256),
num_deconv_kernels=(4, 4),
)
cfg_head_133 = dict(
in_channels=768,
out_channels=133,
num_deconv_layers=2,
num_deconv_filters=(256, 256),
num_deconv_kernels=(4, 4),
)
self.backbone = ViTMoE(**cfg_backbone)
self.keypoint_head = Head(**cfg_head)
self.associate_head = Head(**cfg_head_133)
def forward(self, x):
indices = torch.zeros((x.shape[0]), dtype=torch.long, device=x.device)
back_out = self.backbone(x, indices)
out = self.keypoint_head(back_out)
if True:
indices += 5 # 最后一个是whole body dataset
back_133 = self.backbone(x, indices)
out_133 = self.associate_head(back_133)
out_foot = out_133[:, 17:23]
out = torch.cat([out, out_foot], dim=1)
if False:
import cv2
vis = x[0].permute(1, 2, 0).cpu().numpy()
mean= np.array([0.485, 0.456, 0.406]).reshape(1, 1, 3)
std=np.array([0.229, 0.224, 0.225]).reshape(1, 1 ,3)
vis = np.clip(vis * std + mean, 0., 1.)
vis = (vis[:,:,::-1] * 255).astype(np.uint8)
value = out_133[0].detach().cpu().numpy()
vis_all = []
for i in range(value.shape[0]):
_val = np.clip(value[i], 0., 1.)
_val = (_val * 255).astype(np.uint8)
_val = cv2.resize(_val, None, fx=4, fy=4)
_val = cv2.applyColorMap(_val, cv2.COLORMAP_JET)
_vis = cv2.addWeighted(vis, 0.5, _val, 0.5, 0)
vis_all.append(_vis)
from easymocap.mytools.vis_base import merge
cv2.imwrite('debug.jpg', merge(vis_all))
import ipdb; ipdb.set_trace()
return {
'output': out
}
from ..basetopdown import BaseTopDownModelCache
from ..topdown_keypoints import BaseKeypoints
class MyViT(BaseTopDownModelCache, BaseKeypoints):
def __init__(self, ckpt='data/models/vitpose+_base.pth', single_person=True, url='https://1drv.ms/u/s!AimBgYV7JjTlgcckRZk1bIAuRa_E1w?e=ylDB2G', **kwargs):
super().__init__(name='myvit', bbox_scale=1.25,
res_input=[192, 256], **kwargs)
self.single_person = single_person
model = ComposeVit()
if not os.path.exists(ckpt):
print('')
print('{} not exists, please download it from {} and place it to {}'.format(ckpt, url, ckpt))
print('')
raise FileNotFoundError
ckpt = torch.load(ckpt, map_location='cpu')['state_dict']
ckpt_backbone = {key:val for key, val in ckpt.items() if key.startswith('backbone.')}
ckpt_head = {key:val for key, val in ckpt.items() if key.startswith('keypoint_head.')}
key_whole = 'associate_keypoint_heads.4.'
ckpt_head_133 = {key.replace(key_whole, 'associate_head.'):val for key, val in ckpt.items() if key.startswith(key_whole)}
ckpt_backbone.update(ckpt_head)
ckpt_backbone.update(ckpt_head_133)
state_dict = ckpt_backbone
self.load_checkpoint(model, state_dict, prefix='', strict=True)
model.eval()
self.model = model
self.device = torch.device('cuda') if torch.cuda.is_available() else torch.device('cpu')
self.model.to(self.device)
def dump(self, cachename, output):
_output = output['output']
kpts = self.get_max_preds(_output)
kpts_ori = self.batch_affine_transform(kpts, output['inv_trans'])
kpts = np.concatenate([kpts_ori, kpts[..., -1:]], axis=-1)
output = {'keypoints': kpts}
super().dump(cachename, output)
return output
def estimate_keypoints(self, bbox, images, imgnames):
squeeze = False
if not isinstance(images, list):
images = [images]
imgnames = [imgnames]
bbox = [bbox]
squeeze = True
nViews = len(images)
kpts_all = []
for nv in range(nViews):
_bbox = bbox[nv]
if _bbox.shape[0] == 0:
if self.single_person:
kpts = np.zeros((1, self.num_joints, 3))
else:
kpts = np.zeros((_bbox.shape[0], self.num_joints, 3))
else:
img = images[nv]
# TODO: add flip test
out = super().__call__(_bbox, img, imgnames[nv])
kpts = out['params']['keypoints']
if kpts.shape[-2] == 23:
kpts = self.coco23tobody25(kpts)
elif kpts.shape[-2] == 17:
kpts = self.coco17tobody25(kpts)
else:
raise NotImplementedError
kpts_all.append(kpts)
if self.single_person:
kpts_all = [k[0] for k in kpts_all]
kpts_all = np.stack(kpts_all)
if squeeze:
kpts_all = kpts_all[0]
return {
'keypoints': kpts_all
}
def __call__(self, bbox, images, imgnames):
return self.estimate_keypoints(bbox, images, imgnames)
if __name__ == '__main__':
# Load checkpoint
rand_input = torch.rand(1, 3, 256, 192)
model = MyViT()