Transform pixel location to different reference.

Convert axis-angle representation to rotation matrix. Args: theta: size = [B, 3] Returns: Rotation matrix corresponding to the

Constructs an HMR model with ResNet50 backbone. Args: pretrained (bool): If True, returns a model pre-trained on ImageNet

Constructs an PyMAF model with ResNet50 backbone. Args: pretrained (bool): If True, returns a model pre-trained on ImageNet

Parse input arguments.

Save dictionary state to disk

Crop image according to the supplied bounding box.

Finds bounding box parameters for all keypoints. Look for sequences in the middle with no predictions and linearly interpolate the bbox

This function computes the perspective projection of a set of points. Input: points (bs, N, 3): 3D points rotation (bs, 3, 3)

Convert 6D rotation representation to 3x3 rotation matrix. Based on Zhou et al., "On the Continuity of Rotation Representations in Neural Networks

Rotate all points in a 2D array around the y axis.

Batched version of compute_similarity_transform.

3x3 convolution with padding

Find camera translation that brings 3D joints S closest to 2D the corresponding joints_2d. Input: S: (B, 49, 3) 3D joint locations

Flip keypoints.

flip SMPL pose parameters

example: images / train2017 / 000000119993.jpg

Process gt 2D keypoints and apply all augmentation transforms.

Rotate SMPL pose parameters by rot degrees

This function is borrowed from https://github.com/kornia/kornia Convert 3x4 rotation matrix to Rodrigues vector Args: rotation_

Given 2D points, sample the point-wise features for each point, the dimension of point-wise features will be reduced from C_s to C_p

Save checkpoint.

Training process.

Undo' the image cropping/resizing. This function is used when evaluating mask/part segmentation.

self.root / annotations / person_keypoints_train2017.json

coco ann: [u'segmentation', u'area', u'iscrowd', u'image_id', u'bbox', u'category_id', u'id'] iscrowd: crowd instances ar

ground truth bbox and keypoints

image id: int

Deconv_layer used in Simple Baselines: Xiao et al. Simple Baselines for Human Pose Estimation and Tracking https://github.com

Put results into a dictionary consists of detected people :param trackers (ndarray): input tracklets of shape Nx5 [x1,y1,

Get augmentation parameters.

Computes a similarity transform (sR, t) that takes a set of 3D points S1 (3 x N) closest to a set of 3D points S2, where R is an 3x3 rota

Convert predicted camera from cropped image coordinates to original image coordinates :param cam (ndarray, shape=(3,)): weak perspective

Convert from the class boxes/segms/keyps format generated by the testing code.

Find camera translation that brings 3D joints S closest to 2D the corresponding joints_2d. Input: S: (25, 3) 3D joint locations jo

Check if a checkpoint exists in the current directory.

Flip rgb images or masks. channels come last, e.g. (256,256,3).

Flip pose. The flipping is based on SMPL parameters.

Get a yacs CfgNode object with default values for my_project.