Researchers have developed a new network called MEDN, designed to improve the recognition of micro-expressions by decoupling motion and emotion features. Traditional methods often struggle because micro-expressions can look similar despite conveying opposite emotions. MEDN uses a dual-branch framework to separately process visual motion and implicit emotional cues, with a novel Sparse Emotion Vision Transformer and a Collaborative Fusion Module to combine these distinct features effectively. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
RANK_REASON This is a research paper detailing a novel network architecture for micro-expression recognition.