Human-Object Interaction Detection on HICO-DET Non-rare First v1.0 (test)
25.05mAP (Unseen)GEN-VLKT
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| GEN-VLKTBackbone=ResNet-502023.08 | 25.05 | 23.38 | 23.71 | — | |
| RLIPv2-ParSeDABackbone=ResNet-502023.08 | 22.81 | 29.52 | 28.18 | — | |
| RLIPv2-ParSeDABackbone=Swin-L2023.08 | 22.65 | 40.51 | 36.94 | — | |
| Qpic + SCLK=all, Backbone=ResNet-50, Zero-shot=true2022.03 | 21.73 | 25 | 24.34 | 100 | |
| RLIPv2-ParSeDABackbone=Swin-T2023.08 | 21.07 | 35.07 | 32.27 | — | |
| QpicK=all, Backbone=ResNet-50, Zero-shot=true2022.03 | 21.03 | 23.73 | 23.19 | 100 | |
| RLIPv1-ParSeBackbone=ResNet-502023.08 | 20.27 | 27.67 | 26.19 | — | |
| FCLK=all, Zero-shot=true2022.03 | 18.66 | 19.55 | 19.37 | 100 | |
| FCLBackbone=ResNet-502023.08 | 18.66 | 19.55 | 19.37 | — | |
| ATLK=all, Zero-shot=true2022.03 | 18.25 | 18.78 | 18.67 | 100 | |
| ATLBackbone=ResNet-502023.08 | 18.25 | 18.78 | 18.67 | — | |
| RLIPv1-ParSeDBackbone=ResNet-502023.08 | 16.99 | 24.71 | 22.93 | — | |
| VCLBackbone=ResNet-502023.08 | 16.22 | 18.52 | 18.06 | — | |
| QpicK=360, Backbone=ResNet-50, Zero-shot=true2022.03 | 14.91 | 23.13 | 21.48 | 30.83 | |
| QpicK=120, Backbone=ResNet-50, Zero-shot=true2022.03 | 14.9 | 22.19 | 20.58 | 21.67 | |
| QpicK=240, Backbone=ResNet-50, Zero-shot=true2022.03 | 14.9 | 22.79 | 21.22 | 25 | |
| SCLK=all, Zero-shot=true2022.03 | 13.3 | 16.7 | 16.02 | 100 | |
| SCLK=360, Zero-shot=true2022.03 | 7.91 | 16.7 | 14.94 | 30.83 | |
| SCLK=240, Zero-shot=true2022.03 | 7.17 | 16.7 | 14.8 | 25 | |
| SCLK=120, Zero-shot=true2022.03 | 7.05 | 16.7 | 14.77 | 21.67 | |
| SCLK=0, Zero-shot=true2022.03 | 5.86 | 16.7 | 14.53 | 0 | |
| QpicK=0, Backbone=ResNet-50, Zero-shot=true2022.03 | 0 | 23.73 | 18.98 | 0 |