Action Unit Detection
Action unit (AU) detection, a crucial aspect of facial expression analysis, aims to automatically identify and quantify individual muscle movements in the face, providing a fine-grained representation of emotional states. Current research emphasizes improving AU detection accuracy by addressing challenges like data scarcity and imbalance through techniques such as video masked autoencoders, transformer networks, and multi-modal learning (integrating audio and visual data). These advancements are driven by the need for robust and reliable AU detection in diverse contexts, impacting applications in human-computer interaction, affective computing, and social robotics.
Papers
Multi-label Transformer for Action Unit Detection
Gauthier Tallec, Edouard Yvinec, Arnaud Dapogny, Kevin Bailly
An Attention-based Method for Action Unit Detection at the 3rd ABAW Competition
Duy Le Hoai, Eunchae Lim, Eunbin Choi, Sieun Kim, Sudarshan Pant, Guee-Sang Lee, Soo-Huyng Kim, Hyung-Jeong Yang