Explainable Action Form Assessment by Exploiting Multimodal Chain-of-Thoughts Reasoning
Researchers have introduced a new task called Action Form Assessment (AFA) to evaluate the standardization of human actions and provide corrective feedback. They developed a dataset named CoT-AFA, featuring fitness and martial arts videos with multi-level annotations and a Chain-of-Thought explanation paradigm. The proposed Explainable Fitness Assessor framework uses parallel processing streams and a dynamic gating mechanism to fuse visual and semantic information, improving action classification and quality assessment. AI
IMPACT Introduces a novel approach to action assessment with explainability, potentially improving AI's ability to provide actionable feedback in physical domains.