“Experts propose a 3D hand pose estimation network that integrates hand and object features. Through multi-level cross modal interactive optimization, it effectively improves the performance of hand pose estimation in occluded scenes, opening up new directions for human-computer interaction research in complex interactive scenes.”