Weakly Supervised Action Localization On Gtea
평가 지표
mAP@0.1:0.7
mAP@0.5
평가 결과
이 벤치마크에서 각 모델의 성능 결과
모델 이름 | mAP@0.1:0.7 | mAP@0.5 | Paper Title | Repository |
|---|---|---|---|---|
| AU-Action | 76.9 | 66.3 | Is Weakly-supervised Action Segmentation Ready For Human-Robot Interaction? No, Let's Improve It With Action-union Learning | - |
| HR-Pro | 47.3 | 37.3 | HR-Pro: Point-supervised Temporal Action Localization via Hierarchical Reliability Propagation | |
| SF-Net | 31.0 | 19.3 | SF-Net: Single-Frame Supervision for Temporal Action Localization | |
| LACP | 43.5 | 33.9 | Learning Action Completeness from Points for Weakly-supervised Temporal Action Localization | |
| Li et al. | 36.4 | 28.8 | Temporal Action Segmentation from Timestamp Supervision | |
| Ju et al. | 33.7 | 21.9 | Point-Level Temporal Action Localization: Bridging Fully-supervised Proposals to Weakly-supervised Losses | - |
0 of 6 row(s) selected.