基于单目RGB图像的三维手部姿态估计方法
|
杨冰,徐楚阳,姚金良,向学勤
|
3D hand pose estimation method based on monocular RGB images
|
Bing YANG,Chuyang XU,Jinliang YAO,Xueqin XIANG
|
|
表 1 不同三维手部姿态估计方法在FreiHAND数据集上的性能对比 |
Tab.1 Performance comparison of different 3D hand pose estimation methods in FreiHAND dataset |
|
方法 | PA-MPJPE/mm | PA-MPVPE/mm | F@5 | F@15 | MobileHand[22] | — | 13.1 | 0.439 | 0.902 | FreiHAND[7] | 11.0 | 10.9 | 0.516 | 0.934 | Pose2Mesh[23] | 7.8 | 7.7 | 0.674 | 0.969 | I2L-MeshNet[24] | 7.4 | 7.6 | 0.681 | 0.973 | CMR[15] | 6.9 | 7.0 | 0.715 | 0.977 | I2UV-HandNet[25] | 6.7 | 6.9 | 0.707 | 0.977 | METRO[4] | 6.3 | 6.5 | 0.731 | 0.984 | RealisticHands[26] | — | 7.8 | 0.662 | 0.971 | FastMETRO[5] | 6.5 | 7.2 | 0.688 | 0.983 | PA-Tran[27] | 6.0 | 6.3 | — | — | FastViT-MA36[28] | 6.6 | 6.7 | 0.722 | 0.981 | 本研究(ResNet) | 6.4 | 6.7 | 0.714 | 0.982 | 本研究(HRNet) | 5.8 | 6.2 | 0.749 | 0.986 |
|
|
|