Create sub discriminators.

Return the percentage below the distance threshold, while ignoring distances values with -1. Note: batch_size: N Args: di

Get bbox center

Get bbox center

Get geometric center of all keypoints

Get geometric center of all keypoints

Get keypoint predictions from 3D score maps. Note: batch size: N num keypoints: K heatmap depth size: D heatmap h

Check and initialize input transforms. The in_channels, in_index and input_transform must match. Specifically, when input_transform i

Initialize classification branch and regression branch of head.

Load all image paths and labels from json annotation files into buffer.

Load all image paths and labels from json annotation files into buffer.

Load all image paths and labels from json annotation files into buffer.

Load all image paths and labels from json annotation files into buffer.

Load all image paths and labels from JSON annotation files into buffer.

Load all image paths and labels from JSON annotation files into buffer.

Load all image paths and labels from JSON annotation files into buffer.

Make branches.

Make branches.

Make deconv layers.

Make deconv layers.

Make fuse layer.

Make fuse layer.

Build fuse layer.

Make layer.

Stack blocks to make a layer. Args: out_channels (int): out_channels of the block. num_blocks (int): number of blocks

Make each layer.

Make branches.

Make one branch.

Make one branch.

Make one branch.

Make stem layer.

Make stem layer.

Make transition layer.

Make channel weighting blocks.

Match joints by tags. Use Munkres algorithm to calculate the best match for keypoints grouping. Note: number of keypoints: K

Online hard keypoint mining.

Parse the raw outputs (losses) of the network. Args: losses (dict): Raw output of the network, which usually contain

Apply munkres algorithm to get the best match. Args: scores(np.ndarray): cost matrix. Returns: np.ndarray: best match.

Rescoring mechanism gaussian or linear. Args: overlap: calculated ious scores: target scores. thr: retain oks overlap < t

Resize the feature maps and compute the average. Args: feature_maps (list[Tensor]): Feature maps. align_corners (bool): Align cor

Resize, unsqueeze and concatenate the feature_maps. Args: feature_maps (list[Tensor]): Feature maps. align_corners (bool): Align

Saves module state to `destination` dictionary. This method is modified from :meth:`torch.nn.Module._save_to_state_dict`. Args: modu

Saves module state to `destination` dictionary. This method is modified from :meth:`torch.nn.Module._save_to_state_dict`. Args: module

Scatter the smoothed pose sequences and use them to update the pose results. Args: results (list[list[dict]]): The origin

Distribution aware coordinate decoding method. Note: - heatmap height: H - heatmap width: W Args: heatmap (np.ndarra

Transform inputs for decoder. Args: inputs (list[Tensor] | Tensor): multi-level img features. Returns: Tenso

Add extra files that are required to support MIM into the package. These files will be added by creating a symlink to the originals if the pa

Pad zeros to a compressed keypoints array. Args: compressed_array (np.ndarray): A compressed keypoints array.

Pad zeros to a compressed keypoints array. Args: compressed_array (np.ndarray): A compressed keypoints array.

Pad zeros to a compressed keypoints array. Args: compressed_array (np.ndarray): A compressed keypoints array.

Adjust the coordinates for better accuracy. Note: batch size: N number of keypoints: K heatmap height: H

Adjusts the compatibility of widths and groups. Args: widths (list[int]): Width of each stage. bottleneck_ratio (floa

Allreduce gradients. Args: params (list[torch.Parameters]): List of parameters of a model coalesce (bool, optional): Whether allr

Get the average value of current window.