Name
MoAtt+ST-GAT
Description
This method is one of the a baselines used for evaluation in TRiPOD paper. In this baseline the human pose dynamics forecasting in preformed with an state-of-the-art method abbreviated to Mo-Att [History repeats itself: Human motion prediction via motion attention, Mao et. al., ECCV 2020] and the trajectory forecasting part in performed using [Stgat: Modeling spatial-temporal interactions for human trajectory prediction, Huang et. al., ICCV 2019]. The results are then reported in the global spaces considering both center pose and global trajectory.
Publication title
TRiPOD baseline
Publication authors
-
Publication venue and year
-
Publication URL
Language(s)
Hardware
Website
Source code URL
Creation date
2021-04-16 12:27:33
| Metric | 80/100 ms | 160/240 ms | 320/500 ms | 400/640 ms | 560/900 ms |
|---|---|---|---|---|---|
| PoseTrack VAM | 86.28509054389885 | 100.92061262618158 | 125.31664084265688 | 137.49940700798254 | 147.92333341095946 |
| PoseTrack VIM | 80.60366016728797 | 93.4266681891612 | 115.67968095423143 | 129.53780664462818 | 141.1308742846481 |
| 3DPW VIM | 62.14918563885825 | 97.73990509031908 | 155.2335432042921 | 184.9577965270352 | 250.98034712973086 |