Video Human-Object Interaction Detection on VidHOI
38.65mAP (Overall)SAGE
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| SAGEτa=1, Oracle mode=true2026.07 | 38.65 | — | — | — | 72.44 | 59.22 | 52.22 | 61.96 | |
| ST-Gaze*Oracle mode=true, word embedding module=true2026.07 | 38.46 | — | — | — | 73.62 | 59.16 | 53.76 | 60.57 | |
| SAGEτa=3, Oracle mode=true2026.07 | 38.18 | — | — | — | 72.21 | 59.95 | 52.18 | 62.02 | |
| SAGEτa=5, Oracle mode=true2026.07 | 38.13 | — | — | — | 72.01 | 59.92 | 52.18 | 61.88 | |
| STTranOracle mode=true2026.07 | 28.32 | — | — | — | — | — | — | — | |
| Gaze-TranParadigm=Two-Stage, Detector=YOLOv5 [14]2026.04 | — | 10.4 | 5.46 | 16.83 | — | — | — | — | |
| HOI-DAParadigm=One-Stage2026.04 | — | 16.27 | 12.21 | 22.35 | — | — | — | — | |
| ST-HOIParadigm=Two-Stage, Detector=Faster R-CNN [35]2026.04 | — | 3.1 | 2.1 | 5.9 | — | — | — | — | |
| STTranParadigm=Two-Stage, Detector=Faster R-CNN [35]2026.04 | — | 7.61 | 3.33 | 13.18 | — | — | — | — |