mexca.video.mefarg

Multi-dimensional edge feature-based AU relation graph (MEFARG) learning.

Implementation of the MEFARG model from the paper:

Luo, C., Song, S., Xie, W., Shen, L., Gunes, H. (2022). Learning multi-dimentionsal edge feature-based AU relation graph for facial action unit recognition. arXiv. https://arxiv.org/pdf/2205.01782.pdf

Code adapted from the OpenGraphAU code base (licensed under Apache 2.0).

Module Contents

Classes

MEFARG

Apply a multi-dimensional edge feature-based action unit (AU) relation graph (MEFARG) model.
class mexca.video.mefarg.MEFARG(config: Dict)[source]

Apply a multi-dimensional edge feature-based action unit (AU) relation graph (MEFARG) model.

Predict activations of 27 main and 14 sub AUs from representations of a face image.

Parameters:

config (dict) – Configuration dict of the model with two keys: n_main_aus is the number of main and n_sub_aus is the number of sub AUs to be predicted. If pretrained model weights are loaded, these must match the configuration.

Notes

First generates face representations using a ResNet50 backbone (default weights), then transforms them into AU activations using a multi-dimensional edge feature learning (MEFL) head (see mexca.video.mefl.mefl()).