基于多尺度编码器融合的三维人体姿态估计算法

包晓安,陈恩琳,张娜,涂小妹,吴彪,张庆琪

3D human pose estimation based on multi-scale encoder fusion

Xiaoan BAO,Enlin CHEN,Na ZHANG,Xiaomei TU,Biao WU,Qingqi ZHANG

表 6 选择帧数的消融研究的定量比较结果

Tab.6 Quantitative comparison result of ablation study on selection of frame rate

方法 f kf FLOPs/106 Np v/(帧·s−1) MPJPE/mm Poseformer[13] 27 27 542.1 9.65 428 47.0 StridedTran[14] 81 81 392 4.06 199 45.4 MHFormer[21] 27 27 1031.8 18.92 33 45.9 MSEFN 27 3 409 6.23 470 47.8 MSEFN 27 9 467 7.55 450 45.5 MSEFN 81 3 458 7.11 455 42.8 MSEFN 81 9 521 7.96 449 40.1 MSEFN 81 27 553 9.97 403 39.8