HOIDiffusion enables realistic and diverse 3D hand-object interaction data synthesis through precise structure and appearance control.
Integrating gaze guidance with hand-object interactions for accurate motion prediction.