Spaces:
Running
on
Zero
Running
on
Zero
File size: 6,935 Bytes
a249588 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 |
# Copyright (c) OpenMMLab. All rights reserved.
from collections import OrderedDict
from typing import Tuple
import numpy as np
import torch
from torch import Tensor, nn
from mmpose.evaluation.functional import keypoint_mpjpe
from mmpose.models.utils.tta import flip_coordinates
from mmpose.registry import KEYPOINT_CODECS, MODELS
from mmpose.utils.tensor_utils import to_numpy
from mmpose.utils.typing import (ConfigType, OptConfigType, OptSampleList,
Predictions)
from ..base_head import BaseHead
@MODELS.register_module()
class MotionRegressionHead(BaseHead):
"""Regression head of `MotionBERT`_ by Zhu et al (2022).
Args:
in_channels (int): Number of input channels. Default: 256.
out_channels (int): Number of output channels. Default: 3.
embedding_size (int): Number of embedding channels. Default: 512.
loss (Config): Config for keypoint loss. Defaults to use
:class:`MSELoss`
decoder (Config, optional): The decoder config that controls decoding
keypoint coordinates from the network output. Defaults to ``None``
init_cfg (Config, optional): Config to control the initialization. See
:attr:`default_init_cfg` for default settings
.. _`MotionBERT`: https://arxiv.org/abs/2210.06551
"""
_version = 2
def __init__(self,
in_channels: int = 256,
out_channels: int = 3,
embedding_size: int = 512,
loss: ConfigType = dict(
type='MSELoss', use_target_weight=True),
decoder: OptConfigType = None,
init_cfg: OptConfigType = None):
if init_cfg is None:
init_cfg = self.default_init_cfg
super().__init__(init_cfg)
self.in_channels = in_channels
self.out_channels = out_channels
self.loss_module = MODELS.build(loss)
if decoder is not None:
self.decoder = KEYPOINT_CODECS.build(decoder)
else:
self.decoder = None
# Define fully-connected layers
self.pre_logits = nn.Sequential(
OrderedDict([('fc', nn.Linear(in_channels, embedding_size)),
('act', nn.Tanh())]))
self.fc = nn.Linear(
embedding_size,
out_channels) if embedding_size > 0 else nn.Identity()
def forward(self, feats: Tuple[Tensor]) -> Tensor:
"""Forward the network. The input is multi scale feature maps and the
output is the coordinates.
Args:
feats (Tuple[Tensor]): Multi scale feature maps.
Returns:
Tensor: Output coordinates (and sigmas[optional]).
"""
x = feats # (B, F, K, in_channels)
x = self.pre_logits(x) # (B, F, K, embedding_size)
x = self.fc(x) # (B, F, K, out_channels)
return x
def predict(self,
feats: Tuple[Tensor],
batch_data_samples: OptSampleList,
test_cfg: ConfigType = {}) -> Predictions:
"""Predict results from outputs.
Returns:
preds (sequence[InstanceData]): Prediction results.
Each contains the following fields:
- keypoints: Predicted keypoints of shape (B, N, K, D).
- keypoint_scores: Scores of predicted keypoints of shape
(B, N, K).
"""
if test_cfg.get('flip_test', False):
# TTA: flip test -> feats = [orig, flipped]
assert isinstance(feats, list) and len(feats) == 2
flip_indices = batch_data_samples[0].metainfo['flip_indices']
_feats, _feats_flip = feats
_batch_coords = self.forward(_feats)
_batch_coords_flip = torch.stack([
flip_coordinates(
_batch_coord_flip,
flip_indices=flip_indices,
shift_coords=test_cfg.get('shift_coords', True),
input_size=(1, 1))
for _batch_coord_flip in self.forward(_feats_flip)
],
dim=0)
batch_coords = (_batch_coords + _batch_coords_flip) * 0.5
else:
batch_coords = self.forward(feats)
# Restore global position with camera_param and factor
camera_param = batch_data_samples[0].metainfo.get('camera_param', None)
if camera_param is not None:
w = torch.stack([
torch.from_numpy(np.array([b.metainfo['camera_param']['w']]))
for b in batch_data_samples
])
h = torch.stack([
torch.from_numpy(np.array([b.metainfo['camera_param']['h']]))
for b in batch_data_samples
])
else:
w = torch.stack([
torch.empty((0), dtype=torch.float32)
for _ in batch_data_samples
])
h = torch.stack([
torch.empty((0), dtype=torch.float32)
for _ in batch_data_samples
])
factor = batch_data_samples[0].metainfo.get('factor', None)
if factor is not None:
factor = torch.stack([
torch.from_numpy(b.metainfo['factor'])
for b in batch_data_samples
])
else:
factor = torch.stack([
torch.empty((0), dtype=torch.float32)
for _ in batch_data_samples
])
preds = self.decode((batch_coords, w, h, factor))
return preds
def loss(self,
inputs: Tuple[Tensor],
batch_data_samples: OptSampleList,
train_cfg: ConfigType = {}) -> dict:
"""Calculate losses from a batch of inputs and data samples."""
pred_outputs = self.forward(inputs)
lifting_target_label = torch.stack([
d.gt_instance_labels.lifting_target_label
for d in batch_data_samples
])
lifting_target_weight = torch.stack([
d.gt_instance_labels.lifting_target_weight
for d in batch_data_samples
])
# calculate losses
losses = dict()
loss = self.loss_module(pred_outputs, lifting_target_label,
lifting_target_weight.unsqueeze(-1))
losses.update(loss_pose3d=loss)
# calculate accuracy
mpjpe_err = keypoint_mpjpe(
pred=to_numpy(pred_outputs),
gt=to_numpy(lifting_target_label),
mask=to_numpy(lifting_target_weight) > 0)
mpjpe_pose = torch.tensor(
mpjpe_err, device=lifting_target_label.device)
losses.update(mpjpe=mpjpe_pose)
return losses
@property
def default_init_cfg(self):
init_cfg = [dict(type='TruncNormal', layer=['Linear'], std=0.02)]
return init_cfg
|