深度可分离卷积和人质心引导联合分组的有效人体姿态估计 In this paper, we propose efficient and effective methods for 2D human pose estimation. A new ResBlock is proposed based on depthwise separable convol