Cross-subject Action Unit Detection with Meta Learning and Transformer-based Relation Modeling

TL;DR

This paper introduces a novel meta-learning and transformer-based approach for cross-subject facial Action Unit detection, effectively reducing identity bias and capturing AU relationships to improve accuracy on public datasets.

Contribution

The paper proposes a meta-learning-based AU representation learning method combined with transformer-based relation modeling, enhancing cross-subject AU detection accuracy.

Findings

Improved F1 scores by 1.3% on BP4D dataset

Achieved better generalization across subjects

Outperformed state-of-the-art methods

Abstract

Facial Action Unit (AU) detection is a crucial task for emotion analysis from facial movements. The apparent differences of different subjects sometimes mislead changes brought by AUs, resulting in inaccurate results. However, most of the existing AU detection methods based on deep learning didn't consider the identity information of different subjects. The paper proposes a meta-learning-based cross-subject AU detection model to eliminate the identity-caused differences. Besides, a transformer-based relation learning module is introduced to learn the latent relations of multiple AUs. To be specific, our proposed work is composed of two sub-tasks. The first sub-task is meta-learning-based AU local region representation learning, called MARL, which learns discriminative representation of local AU regions that incorporates the shared information of multiple subjects and eliminates identity-caused differences. The second sub-task uses the local region representation of AU of the first sub-task as input, then adds relationship learning based on the transformer encoder architecture to capture AU relationships. The entire training process is cascaded. Ablation study and visualization show that our MARL can eliminate identity-caused differences, thus obtaining a robust and generalized AU discriminative embedding representation. Our results prove that on the two public datasets BP4D and DISFA, our method is superior to the state-of-the-art technology, and the F1 score is improved by 1.3% and 1.4%, respectively.