ResLayer to build ResNet style backbone. Args: block (nn.Module): Residual block used to build ResLayer. num_blocks (int): Number

Basic 3d convolution block for I3D. Args: in_channels (int): Input channels of this block. out_channels (int): Output channels of this bl

Inception block for I3D. Args: in_channels (int): Input channels of this block. out_channels (int): Output channels of this block. ex

A residual 3D convolutional block. Args: in_channels (int): Input channels of this block. out_channels (int): Output channels of

Generation of cropped input person, pose coords, smpl parameters. Parameters ---------- img: torch.Tensor A tensor with shape: `(

Generation of cropped input person, pose coords, smpl parameters. Parameters ---------- img: torch.Tensor A tensor with shape: `(

Maintains (exponential) moving average of a set of parameters.

Image to Patch Embedding

A simple timer.

A custom implementation for SMPLer_X pipeline Need to run tool/cache/fix_cache.py to fix paths

MPI-INF-3DHP dataset. Parameters ---------- ann_file: str, Path to the annotation json file. root: str, default './data/3dhp

InvertedResidual block for MobileNetV2. Args: in_channels (int): The input channels of the InvertedResidual block. out_channels (

MPII Person dataset. Parameters ---------- ann_file: str, Path to the annotation json file. root: str, default './data/mpii'

COCO Person dataset. Parameters ---------- ann_file: str, Path to the annotation json file. root: str, default './data/coco'

UP3D Person dataset. Parameters ---------- ann_file: str, Path to the annotation json file. root: str, default './data/up3d'

Applies padding to input (if needed) so that input can get fully covered by filter you specified. It support two modes "same" and "corner". The

Modified Hourglass Module for HourglassNet_AE backbone. Generate module recursively and use BasicBlock as the base unit. Args: depth

Hourglass Module for HourglassNet backbone. Generate module recursively and use BasicBlock as the base unit. Args: depth (int): Dept

A 3D max-pool block. Args: pool_size (int): Pool size of the 3D max-pool layer

Squeeze-and-Excitation Module. Args: channels (int): The input (and output) channels of the SE layer. ratio (int): Squeeze ratio

InvertedResidual block for ShuffleNetV2 backbone. Args: in_channels (int): The input channels of the block. out_channels (int): T

Transformer block for dynamic tokens. Args: dim (int): The feature dimension. num_heads (int): Parallel attention heads.

Conv layer for dynamic tokens. A skip link is added between the input and output tokens to reserve detail tokens.

A 3D upsample block. Args: in_channels (int): Input channels of this block. out_channels (int): Output channels of this block.

Associative Embedding loss. `Associative Embedding: End-to-End Learning for Joint Detection and Grouping <https://arxiv.org/abs/1611.05424v2>

An implementation of the absolute position embedding in PVT. Args: pos_shape (int): The shape of the absolute position embedding.

A basic 3D convolutional block. Args: in_channels (int): Input channels of this block. out_channels (int): Output channels of thi

Basic block for VideoPose3D. Args: in_channels (int): Input channels of this block. out_channels (int): Output channels of this b

Bone length loss. Args: joint_parents (list): Indices of each joint's parent joint. use_target_weight (bool): Option to use weigh

Clustering-based Token Merging module in TCFormer. Args: sample_ratio (float): The sample ratio of tokens. embed_dim (int): Input

A custom implementation for SMPLer_X pipeline Need to run tool/cache/fix_cache.py to fix paths

A custom implementation for SMPLer_X pipeline Need to run tool/cache/fix_cache.py to fix paths

A custom implementation for SMPLer_X pipeline Need to run tool/cache/fix_cache.py to fix paths

A custom implementation for SMPLer_X pipeline Need to run tool/cache/fix_cache.py to fix paths

A custom implementation for SMPLer_X pipeline Need to run tool/cache/fix_cache.py to fix paths

Conditional channel weighting block. Args: in_channels (int): The input channels of the block. stride (int): Stride of the 3x3 co

CpmBlock for Convolutional Pose Machine. Args: in_channels (int): Input channels of this block. channels (list): Output channels

r"""FFN with Depthwise Conv of HRFormer. Args: in_features (int): The feature dimension. hidden_features (int, optional): The hid

Cross-resolution channel weighting module. Args: channels (int): The channels of the module. ratio (int): channel reduction ratio

Depthwise Conv for regular grid-based tokens. Args: dim (int): The feature dimension.

Downsample module for MSPN. Args: block (nn.Module): Downsample block. num_blocks (list): Number of blocks in each downsample uni

Downsample module for RSN. Args: block (nn.Module): Downsample block. num_blocks (list): Number of blocks in each downsample unit

Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks).

Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks).

An encoder-decoder block. Args: in_channels (int): Input channels of this block

Discriminator for SMPL pose parameters of all joints. Args: fc_layers (Tuple): Tuple of neuron count, such as (736, 1024, 102

Human3.6M smpl dataset. 17 Human3.6M joints + 29 SMPL joints Parameters ---------- ann_file: str, Path to the annotation json fi

Human3.6M smpl dataset. 17 Human3.6M joints + 29 SMPL joints Parameters ---------- ann_file: str, Path to the annotation json fi

High-Resolution Module for HRFormer. Args: num_branches (int): The number of branches in the HRFormerModule. block (nn.Module): T

High-Resolution Module for HRNet. In this module, every branch has 4 BasicBlocks/Bottlenecks. Fusion/Exchange is in this module.

Accumulate the heatmap loss for each image in the batch. Args: supervise_empty (bool): Whether to supervise empty channels.