Rethinking the Learning Paradigm for Facial Expression Recognition

TL;DR

This paper challenges the traditional supervised learning approach for facial expression recognition by proposing the use of weakly supervised strategies to better handle ambiguous annotations inherent in real-world datasets.

Contribution

It introduces a novel training paradigm that leverages weak supervision to improve FER model performance on ambiguous, real-world data.

Findings

Weakly supervised training improves FER accuracy.

Handling ambiguous annotations enhances model robustness.

Proposed method outperforms traditional supervised approaches.

Abstract

Due to the subjective crowdsourcing annotations and the inherent inter-class similarity of facial expressions, the real-world Facial Expression Recognition (FER) datasets usually exhibit ambiguous annotation. To simplify the learning paradigm, most previous methods convert ambiguous annotation results into precise one-hot annotations and train FER models in an end-to-end supervised manner. In this paper, we rethink the existing training paradigm and propose that it is better to use weakly supervised strategies to train FER models with original ambiguous annotation.