Abstract: Recognising and tracking human actions from videos is crucial for human-robot collaborative assembly (HRCA). However, traditional action segmentation methods suffer from limited scene ...