Abstract: Transformer-based methods have shown promising performance in 2D-to-3D human pose estimation. However, existing approaches often ignore the relationships between specific joints of the human ...