Human-Object Interaction Detection on HICO-DET computed-box setting
9.35mAP (Full)Qwen2-VL 7B + AnyHOI
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen2-VL 7B + AnyHOIAnyHOI Augmentation=true2026.04 | 9.35 | 10.62 | 8.97 | 32.92 | |
| LLaVA OV 7B + AnyHOIAnyHOI Augmentation=true2026.04 | 8.73 | 10.56 | 8.19 | 28.59 | |
| CogVLM2 + AnyHOIAnyHOI Augmentation=true2026.04 | 8.1 | 7.43 | 8.31 | 29.17 | |
| LAINVocabulary Source=VG2026.04 | 6.97 | 2.72 | 8.24 | 30.92 | |
| MP-HOIVocabulary Source=VG2026.04 | 6.75 | 11.28 | 5.4 | 29.85 | |
| InternVL2 8B + AnyHOIAnyHOI Augmentation=true2026.04 | 6.47 | 5.79 | 6.68 | 26.84 | |
| LLaVA OV 0.5B + AnyHOIAnyHOI Augmentation=true2026.04 | 6.43 | 5.09 | 6.84 | 20.04 | |
| MP-HOIVocabulary Source=WN2026.04 | 5.47 | 8.25 | 4.64 | 27.71 | |
| GEN-VLKTVocabulary Source=VG2026.04 | 4.99 | 5.18 | 4.93 | 29.77 | |
| Qwen2-VL 7BAnyHOI Augmentation=false2026.04 | 4.03 | 4.54 | 3.87 | 11.65 | |
| GEN-VLKTVocabulary Source=WN2026.04 | 3.67 | 2.09 | 4.14 | 28.89 | |
| CLIPVocabulary Source=VG2026.04 | 3.44 | 7.42 | 2.26 | 26.46 | |
| CLIPVocabulary Source=WN2026.04 | 3.13 | 7.9 | 1.7 | 23.57 | |
| HOICLIPVocabulary Source=VG2026.04 | 2.76 | 2.27 | 2.9 | 26.86 | |
| LLaVA OV 7BAnyHOI Augmentation=false2026.04 | 2.72 | 2.49 | 2.79 | 8.05 | |
| LogicHOIVocabulary Source=VG2026.04 | 2.16 | 0.87 | 2.55 | 20.96 | |
| InternVL2 8BAnyHOI Augmentation=false2026.04 | 1.91 | 1.2 | 2.12 | 8.06 | |
| HOICLIPVocabulary Source=WN2026.04 | 1.52 | 2.15 | 1.34 | 21.21 | |
| LogicHOIVocabulary Source=WN2026.04 | 1.49 | 0.47 | 1.79 | 19.23 | |
| CogVLM2AnyHOI Augmentation=false2026.04 | 1.34 | 0.67 | 1.54 | 7.12 | |
| LLaVA OV 0.5BAnyHOI Augmentation=false2026.04 | 1.32 | 0.19 | 1.66 | 5.73 | |
| LAINVocabulary Source=WN2026.04 | 1.05 | 1.05 | 1.05 | 22.83 | |
| Random2026.04 | 0.69 | 0.27 | 0.81 | 22.22 |