Abstract: Despite the remarkable progress in 3D human pose estimation, existing methods face dual challenges of insufficient spatiotemporal modeling and inefficient cross-modal interaction in complex ...