UAU-Net: Uncertainty-aware Representation Learning and Evidential Classification for Facial Action Unit Detection

TL;DR

UAU-Net introduces a novel framework for facial action unit detection that explicitly models uncertainty at both feature extraction and classification stages, enhancing robustness and reliability.

Contribution

The paper presents UAU-Net, combining CV-AFE for probabilistic feature learning and AB-ENN for uncertainty-aware classification, addressing limitations of deterministic models.

Findings

Improves AU detection accuracy on BP4D and DISFA datasets.

Models uncertainty to enhance robustness against noise and label imbalance.

Reduces overconfidence in predictions through evidential neural networks.

Abstract

Facial action unit (AU) detection remains challenging because it involves heterogeneous, AU-specific uncertainties arising at both the representation and decision stages. Recent methods have improved discriminative feature learning, but they often treat the AU representations as deterministic, overlooking uncertainty caused by visual noise, subject-dependent appearance variations, and ambiguous inter-AU relationships, all of which can substantially degrade robustness. Meanwhile, conventional point-estimation classifiers often provide poorly calibrated confidence, producing overconfident predictions, especially under the severe label imbalance typical of AU datasets. We propose UAU-Net, an Uncertainty-aware AU detection framework that explicitly models uncertainty at both stages. At the representation stage, we introduce CV-AFE, a conditional VAE (CVAE)-based AU feature extraction module that learns probabilistic AU representations by jointly estimating feature means and variances across multiple spatio-temporal scales; conditioning on AU labels further enables CV-AFE to capture uncertainty associated with inter-AU dependencies. At the decision stage, we design AB-ENN, an Asymmetric Beta Evidential Neural Network for multi-label AU detection, which parameterizes predictive uncertainty with Beta distributions and mitigates overconfidence via an asymmetric loss tailored to highly imbalanced binary labels. Extensive experiments on BP4D and DISFA show that UAU-Net achieves strong AU detection performance, and further analyses indicate that modeling uncertainty in both representation learning and evidential prediction improves robustness and reliability.