Create a NODE and add it to the node_list. name, idx, parent respectively corespond to the "name", "index", "parent" keys.

@ bone_length: (batch_size, N_joints, 1) or (1, N_joints, 1)

create human tree structure according to data_type return a DictTree object.

Converts homogeneous points to euclidean Args: points numpy array or torch tensor of shape (N, M + 1): N homogeneous points of dimension

Implementation of Lagrangian Algorithm for constrained HPE problem.

Reverse to normalize_image() function

Visualizes a 2d skeleton Args keypoints numpy array of shape (19, 2): pose to draw in CMU format. ax: matplotlib axis to dra

:param center: (2,) :param scale: (2,) :param patch_size: :param inv: inv=0 image->crop_img :return:

Project 3D points to image plane not taking into account distortion Args: proj_matrix numpy array or torch tensor of shape (3, 4): project

Structural Triangulation with step contrain algorithm. When n_step == 1, this is pure ST without SCA.

Return an activation function given a string

Computes the precision@k for the specified values of k

:param pts: (2, N) :param trans: (2, 3) :return:

3x3 convolution with padding

3x3 convolution with padding

Converts euclidean points to homogeneous Args: points numpy array or torch tensor of shape (N, M): N euclidean points of dimension M

Nearby joints NMS implementations. Args: kpts_db (list[dict]): keypoints and scores. dist_thr (float): threshold for judging whet