Temporal Action Detection on FineAction
29.4Avg mAPOur MambaTAD
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Our MambaTADFeatures=IV1B2025.11 | 29.4 | 45.6 | 29.6 | 5.8 | |
| VideoMambaSuiteFeatures=IV1B2025.11 | 29 | 45.4 | 28.8 | 6.8 | |
| InternVideo2s1-6BBackbone=InternVideo2s1-6B2024.03 | 27.7 | — | — | — | |
| InternVideo2Features=IV6B2025.11 | 27.7 | — | — | — | |
| InternVideo2s1-1BBackbone=InternVideo2s1-1B2024.03 | 27.2 | — | — | — | |
| ActionFormer*Features=IV6B2025.11 | 27.2 | 43.1 | 27.1 | 5.32 | |
| InternVideo2Features=IV1B2025.11 | 27.2 | — | — | — | |
| DyFADetBackbone=VM2-g2024.07 | 23.8 | 37.1 | 23.7 | 5.9 | |
| DyFADetFeatures=VM2-g2025.11 | 23.8 | 37.1 | 23.7 | 5.9 | |
| LFAFFeatures=X-CLIP2025.11 | 22.2 | 36.9 | 21.3 | 4.5 | |
| ActionFormerBackbone=VM2-g2024.07 | 18.2 | 29.1 | 17.7 | 5.1 | |
| VideoMAEv2-gBackbone=VideoMAEv2-g2024.03 | 18.2 | — | — | — | |
| ActionFormerFeatures=VM2-g2025.11 | 18.2 | 29.1 | 17.7 | 5.1 | |
| ActionFormerBackbone=InternVideo2024.07 | 17.6 | — | — | — | |
| InternVideoBackbone=InternVideo2024.03 | 17.6 | — | — | — | |
| InternVideoFeatures=IV2025.11 | 17.6 | — | — | — | |
| ViT-TADFeatures=ViT-B2025.11 | 17.2 | 32.6 | 15.9 | 2.7 | |
| PlusTAD (Anchor-based)Backbone=R50-SlowOnly, RGB-Only=true2022.05 | 12.15 | 24.34 | 10.57 | 0.43 | |
| PlusTAD (Anchor-free)Backbone=R50-SlowOnly, RGB-Only=true2022.05 | 11.66 | 22.37 | 10.36 | 0.83 | |
| BMNBackbone=I3D2024.07 | 9.3 | 14.4 | 8.9 | 3.1 | |
| BMNSupervision=fully-supervised2022.07 | 9.25 | 14.44 | 8.92 | 3.12 | |
| BMNBackbone=I3D, RGB-Only=false2022.05 | 9.25 | 14.44 | 8.92 | 3.12 | |
| G-TADBackbone=I3D2024.07 | 9.1 | 13.7 | 8.8 | 3.1 | |
| G-TADBackbone=I3D, RGB-Only=false2022.05 | 9.06 | 13.74 | 8.83 | 3.06 | |
| BMNBackbone=I3D, RGB-Only=true2022.05 | 7.86 | 12.56 | 7.49 | 2.62 | |
| G-TADBackbone=I3D, RGB-Only=true2022.05 | 6.87 | 10.88 | 6.52 | 2.19 | |
| DBGBackbone=I3D, RGB-Only=false2022.05 | 6.75 | 10.65 | 6.43 | 2.5 | |
| DBGBackbone=I3D, RGB-Only=true2022.05 | 5.31 | 8.57 | 5.01 | 1.93 | |
| HAANLoss components=L_mil + L_pseudo + L_concept + L_coarse, Supervision=weakly-supervised2022.07 | 4.1 | 7.05 | 3.95 | 1.14 | |
| WTAL-UMSupervision=weakly-supervised2022.07 | 3.64 | 6.65 | 3.23 | 0.95 | |
| HAANLoss components=L_mil + L_pseudo + L_concept, Supervision=weakly-supervised2022.07 | 3.57 | 6.57 | 3.27 | 0.83 | |
| W-TALCSupervision=weakly-supervised2022.07 | 3.45 | 6.18 | 3.15 | 0.83 | |
| D2-NetSupervision=weakly-supervised2022.07 | 3.35 | 6.75 | 3.02 | 0.82 | |
| ASLSupervision=weakly-supervised2022.07 | 3.3 | 6.79 | 2.68 | 0.81 | |
| HAANLoss components=L_mil + L_pseudo, Supervision=weakly-supervised2022.07 | 3.18 | 6.4 | 2.73 | 0.61 | |
| RefineLocSupervision=weakly-supervised2022.07 | 3.02 | 5.93 | 2.55 | 0.96 | |
| HAANLoss components=L_mil, Supervision=weakly-supervised2022.07 | 2.72 | 5.74 | 2.29 | 0.39 |