Action Recognition on NTU 120 (test)
63.2AccuracySkeleton Transformer
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Skeleton TransformerZero-shot Transfer=true2026.03 | 63.2 | — | — | — | |
| ST-GCNZero-shot Transfer=true2026.03 | 58.9 | — | — | — | |
| Trans4SOAR (Small)scenario=Random Occlusion, model_scale=Small2022.02 | 53.27 | 51.33 | 53.8 | 53.35 | |
| Trans4SOAR (Base)scenario=Random Occlusion, model_scale=Base2022.02 | 53.17 | 52.89 | 54.5 | 53.21 | |
| SL-DML (Swin)scenario=Random Occlusion, backbone=Swin2022.02 | 47.19 | 45.64 | 46.78 | 47.29 | |
| SL-DMLscenario=Random Occlusion2022.02 | 42.53 | 42.24 | 44.79 | 42.56 | |
| SL-DML (LeViT) + LSCscenario=Random Occlusion, backbone=LeViT, loss=LSC2022.02 | 38.04 | 35.93 | 37.87 | 38.11 | |
| Skeleton-DMLscenario=Random Occlusion2022.02 | 35.15 | 32.59 | 34.29 | 35.22 | |
| SL-DML (LeViT)scenario=Random Occlusion, backbone=LeViT2022.02 | 35 | 33.24 | 41.45 | 35.1 |