data class for loading AFLW2000 dataset make sure each image has corresponding mat file, which provides cropping infromation

https://github.com/timesler/facenet-pytorch/blob/master/examples/infer.ipynb

testpath: folder, imagepath_list, image path, video path

# 53877 faces K must be less than 6

# 53877 faces K must be less than 6

testpath: folder, imagepath_list, image path, video path

testpath: folder, imagepath_list, image path, video path

K must be less than 6

K must be less than 6

K must be less than 6

Calculates landmarks by barycentric interpolation Input: vertices: torch.tensor NxVx3, dtype = torch.float32

for similarity of neighbors ref: Self-supervised Multi-level Face Model Learning for Monocular Reconstruction at over 250 Hz Towards

computes ring loss for ring_outputs before FLAME decoder Inputs: ring_outputs = a list containing N streams of

computes ring loss for ring_outputs before FLAME decoder Inputs: ring_outputs = a list containing N streams of

get rotation matrix from three rotation angles(degree). right-handed. Args: angles: [batch_size, 3] tensor containing X, Y, and Z angles.

r""" Returns: `dict` of attention processors: A dictionary containing all attention processors used in the model with

X is N x num_pquaternion_to_angle_axisoints x 3

orthgraphic projection X: 3d vertices, [bz, n_point, 3] camera: scale and translation, [bz, 3], [scale, tx, ty]

Function that converts angles from degrees to radians. See :class:`~torchgeometry.DegToRad` for details. Args: tensor (Tensor): Tens

Disables the FreeU mechanism.

r"""Enables the FreeU mechanism from https://arxiv.org/abs/2309.11497. The suffixes after the scaling factors represent the stage blocks wher

:param vertices: [batch size, number of vertices, 3] :param faces: [batch size, number of faces, 3] :return: [batch size, number of face

Compute the faces, barycentric coordinates for the dynamic landmarks To do so, we first compute the rotation of the neck around the y-axis

purpose: flip a image given by src_image and the 2d keypoints flip_mode: 0: horizontal flip >0: