Global Human Mesh Recovery on Dynamic Human3.6M 1.0 (Invisible frames)
8.2FIDGLAMR
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GLAMRBase Mesh Recovery Method=SPEC2021.12 | 8.2 | 72.8 | |
| GLAMRBase Mesh Recovery Method=KAMA2021.12 | 11.4 | 67.7 | |
| SPECBase Mesh Recovery Method=SPEC, Motion Infilling Strategy=ConvAE2021.12 | 28.5 | 80.1 | |
| KAMABase Mesh Recovery Method=KAMA, Motion Infilling Strategy=ConvAE2021.12 | 28.9 | 77.4 | |
| SPECBase Mesh Recovery Method=SPEC, Motion Infilling Strategy=Linear Interpolation2021.12 | 29.7 | 78.7 | |
| KAMABase Mesh Recovery Method=KAMA, Motion Infilling Strategy=Linear Interpolation2021.12 | 30.2 | 74.8 | |
| SPECBase Mesh Recovery Method=SPEC, Motion Infilling Strategy=Last Pose2021.12 | 36.2 | 92.6 | |
| KAMABase Mesh Recovery Method=KAMA, Motion Infilling Strategy=Last Pose2021.12 | 36.7 | 88.8 |