Uncertain Graph Neural Networks for Facial Action Unit Detection

Capturing the dependencies among different facial action units (AU) is extremely important for AU detection task. Many studies have employed graph-based deep learning methods to exploit AUs. However, AUs in real world data are often noisy and uncertainty essential be taken into consideration. Rather than employing a deterministic mode, we propose an uncertain graph neural network (UGN) learn probabilistic mask that simultaneously captures both individual uncertainties. Further, adaptive weighted loss function based on epistemic uncertainties adaptively vary weights of training samples during process account unbalanced distributions We also provide insightful analysis how related performance detection. Extensive experiments, conducted two benchmark datasets, i.e., BP4D DISFA, demonstrate our method achieves state-of-the-art performance.