Returns the angle prior loss for the given pose Args: pose: (Bx[23 + 1] * 3) torch tensor with the axis-angle repres

Performs the matching Params: outputs: This is a dict that contains at least these entries: "pred_logits": Tens

This performs the loss computation. Parameters: outputs: dict of tensors, see the output specification of the model

Forward functions for Transformer Network Input: @ views: a list of images with shape [batch_size, 3, H, W]

During each decoder layer, each 3D poses of each queries will be projected into each camera view to aggregate features and update coa

A decoder forward function, that contains multiple decoder layers. Most of the the important stuff is written in DQDecoderLay

:param query (n_views, Length_{query}, C) :param reference_points (n_views, Length_{query}, n_levels, 2

Generalized IoU from https://giou.stanford.edu/ The boxes should be in [x0, y0, x1, y1] format Returns a [N, M] pairwise matrix, where

:param joints: [[num_joints, 3]] :param joints_vis: [num_joints, 3] :return: input_heatmap

:param joints: [[num_joints, 3]] :param joints_vis: [num_joints, 3] :return: target, target_weight(1: visible, 0: invisible)

:pose3D: <numpy.ndarray> of n_frames x n_joints x 3, the 3D joint coordinates. :return: <numpy.ndarray> of n_frames x n_bones, the 3D

Returns the mean of the mixture

Get ray origin and normalized directions in world coordinate for all pixels in one image. Reference: https://www.scratchapixel.com/lesson

Makes square bbox from any bbox by stretching of minimal length side Args: bbox tuple of size 4: input bbox (left, upper, right, lower)

Applies softmax to heatmaps and integrates them to get their's "center of masses" Args: heatmaps torch tensor of shape (batch_size, n_hea

Equivalent to nn.functional.interpolate, but with support for empty batch sizes. This will eventually be supported natively by PyTorch, a

Equivalent to nn.functional.interpolate, but with support for empty batch sizes. This will eventually be supported natively by PyTorch, a

Compute the cardinality error, ie the absolute error in the number of predicted non-empty poses This is not really a loss, it is inte

Classification loss (NLL) targets dicts must contain the key "labels" containing a tensor of dim [nb_target_poses]

Compute the bounding boxes around the provided masks The masks should be in format [N, H, W] where N is the number of masks, (H, W) are the s

Greedily select boxes with high confidence and overlap <= thr. Args: dets: [[x1, y1, x2, y2, score]]. thr: Retain overlap < thr.

OKS NMS implementations. Args: kpts_db: keypoints. thr: Retain overlap < thr. sigmas: standard deviation of keypoint labe

Perform orthographic projection of 3D points X using the camera parameters Args: X: size = [B, N, 3] camera: size = [B, 3] Ret