Human-Object Interaction Detection on HICO-DET
50.56mAP (Full)HOI-IDiff
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| HOI-IDiffEvaluation setting=Known objects setting2025.03 | 50.56 | — | — | — | — | — | — | 51.95 | 50.14 | — | |
| Wu et al.Evaluation setting=Known objects setting2025.03 | 49.5 | — | — | — | — | — | — | 50.59 | 49.18 | — | |
| Luo et al.Evaluation setting=Known objects setting2025.03 | 48.16 | — | — | — | — | — | — | 48.37 | 48.09 | — | |
| GRASP-HOIBackbone=ResNet-50+MLLM, Closed-set setting=true2025.12 | 48.02 | — | — | — | — | — | — | 48.15 | 48.09 | — | |
| Zhang et al. [63]Evaluation setting=Known objects setting2025.03 | 47.81 | — | — | — | — | — | — | 48.38 | 47.64 | — | |
| HOI-IDiffEvaluation setting=Default setting2025.03 | 47.71 | — | — | — | — | — | — | 48.36 | 47.52 | — | |
| HORPBackbone=Swin-L+CLIP, Closed-set setting=true2025.12 | 47.53 | — | — | — | — | — | — | 46.81 | 47.74 | — | |
| DA-HOIDetector=Grounding-DINO, Supervision Setting=fully supervised2026.02 | 46.21 | — | — | — | — | — | — | 47.54 | 45.81 | — | |
| Wu et al.Evaluation setting=Default setting2025.03 | 46.01 | — | — | — | — | — | — | 46.74 | 45.8 | — | |
| PAFRBackbone=Swin-L, Closed-set setting=true2025.12 | 46.01 | — | — | — | — | — | — | 46.74 | 45.8 | — | |
| DA-HOIDetector=Yolo-World, Supervision Setting=fully supervised2026.02 | 45.62 | — | — | — | — | — | — | 49.34 | 44.51 | — | |
| Yuan et al.Evaluation setting=Default setting2025.03 | 45.09 | — | — | — | — | — | — | 43.23 | 45.64 | — | |
| RLIPv2Backbone=Swin-L, Closed-set setting=true2025.12 | 45.09 | — | — | — | — | — | — | 43.23 | 45.64 | — | |
| Luo et al.Evaluation setting=Default setting2025.03 | 45.04 | — | — | — | — | — | — | 45.61 | 44.88 | — | |
| SICHOIBackbone=R101+VIT-L, Closed-set setting=true2025.12 | 45.04 | — | — | — | — | — | — | 45.61 | 44.88 | — | |
| DA-HOISupervision Setting=fully supervised2026.02 | 44.58 | — | — | — | — | — | — | 46.17 | 44.08 | — | |
| Yang et al.Evaluation setting=Default setting2025.03 | 44.53 | — | — | — | — | — | — | 44.48 | 44.55 | — | |
| MP-HOIBackbone=Swin-L+ViT, Closed-set setting=true2025.12 | 44.53 | — | — | — | — | — | — | 44.48 | 44.55 | — | |
| Zhang et al. [63]Evaluation setting=Default setting2025.03 | 44.32 | — | — | — | — | — | — | 44.61 | 44.24 | — | |
| PVICBackbone=Swin-L, Closed-set setting=true2025.12 | 44.32 | — | — | — | — | — | — | 44.61 | 44.24 | — | |
| BC-HOIBackbone=ResNet-50, Closed-set setting=true2025.12 | 43.01 | — | — | — | — | — | — | 45.76 | 42.18 | — | |
| BC-HOIFeature Extractor=BLIP2, Supervision Setting=fully supervised2026.02 | 43.01 | — | — | — | — | — | — | 45.76 | 42.18 | — | |
| Wang et al.Evaluation setting=Known objects setting2025.03 | 42.24 | — | — | — | — | — | — | 42.86 | 42.05 | — | |
| DiffHOIBackbone=Swin-L, Closed-set setting=true2025.12 | 41.5 | — | — | — | — | — | — | 39.96 | 41.96 | — | |
| UniHOI-l (w/ BLIP2)Extras=L2023.11 | 40.95 | — | — | — | — | — | — | 40.27 | 41.32 | — | |
| UniHOI-1Backbone=R101+VIT-L, Closed-set setting=true2025.12 | 40.95 | — | — | — | — | — | — | 40.27 | 41.32 | — | |
| UniHOI-m (w/ BLIP2)Extras=L2023.11 | 40.74 | — | — | — | — | — | — | 40.03 | 40.95 | — | |
| UniHOI-s (w/ BLIP2)Extras=L2023.11 | 40.06 | — | — | — | — | — | — | 39.91 | 40.11 | — | |
| UniHOIFeature Extractor=BLIP2, Supervision Setting=fully supervised2026.02 | 40.06 | — | — | — | — | — | — | 39.91 | 40.11 | — | |
| Wang et al.Evaluation setting=Default setting2025.03 | 39.34 | — | — | — | — | — | — | 39.9 | 39.17 | — | |
| BCOMBackbone=ResNet-50, Closed-set setting=true2025.12 | 39.34 | — | — | — | — | — | — | 39.9 | 39.17 | — | |
| BCOMSupervision Setting=fully supervised2026.02 | 39.34 | — | — | — | — | — | — | 39.9 | 39.17 | — | |
| Ma et al.Evaluation setting=Known objects setting2025.03 | 38.93 | — | — | — | — | — | — | 31.93 | 41.02 | — | |
| GEN-VLKT-L + Category Query LearningDetector=HICO-DET, Backbone=ResNet101, Evaluation Setting=Known Object2023.03 | 38.82 | — | — | — | — | — | — | 35.51 | 39.81 | — | |
| GEN-VLKT-M + Category Query LearningDetector=HICO-DET, Backbone=ResNet101, Evaluation Setting=Known Object2023.03 | 38.79 | — | — | — | — | — | — | 35.28 | 39.84 | — | |
| EZ-HOIBackbone=ResNet-50, Closed-set setting=true2025.12 | 38.61 | — | — | — | — | — | — | 37.7 | 38.89 | — | |
| EZ-HOISupervision Setting=fully supervised2026.02 | 38.61 | — | — | — | — | — | — | 37.7 | 38.89 | — | |
| GEN-VLKT-B + Category Query LearningDetector=HICO-DET, Backbone=ResNet50, Evaluation Setting=Known Object2023.03 | 38.43 | — | — | — | — | — | — | 34.85 | 39.5 | — | |
| ADA-CMBackbone=R50+CLIP, Closed-set setting=true2025.12 | 38.4 | — | — | — | — | — | — | 37.52 | 38.66 | — | |
| Liao et al.Evaluation setting=Known objects setting2025.03 | 38.22 | — | — | — | — | — | — | 34.94 | 39.37 | — | |
| GEN-VLKT-LDetector=HICO-DET, Backbone=ResNet101, Evaluation Setting=Known Object2023.03 | 38.22 | — | — | — | — | — | — | 34.36 | 39.37 | — | |
| CMMPBackbone=ResNet-50, Closed-set setting=true2025.12 | 38.14 | — | — | — | — | — | — | 37.75 | 38.25 | — | |
| CMMPSupervision Setting=fully supervised2026.02 | 38.14 | — | — | — | — | — | — | 37.75 | 38.25 | — | |
| RegFormerSupervision=Weakly Supervised, Detector=H-DETR, Vision Backbone=DINO-B, Text Backbone=CLIP-B2026.04 | 38.14 | — | — | — | — | — | — | 40.31 | 37.49 | — | |
| GEN-VLKT-MDetector=HICO-DET, Backbone=ResNet101, Evaluation Setting=Known Object2023.03 | 38.07 | — | — | — | — | — | — | 34.94 | 39.01 | — | |
| Chen et al.Evaluation setting=Known objects setting2025.03 | 37.59 | — | — | — | — | — | — | 31.36 | 39.36 | — | |
| BodyPartMapDetector=HICO-DET, Backbone=ResNet50, Evaluation Setting=Known Object2023.03 | 37.56 | — | — | — | — | — | — | 35.87 | 38.06 | — | |
| RmLRBackbone=ResNet-101, Closed-set setting=true2025.12 | 37.41 | — | — | — | — | — | — | 28.81 | 39.97 | — | |
| VIPLOBackbone=R50+CLIP, Closed-set setting=true2025.12 | 37.22 | — | — | — | — | — | — | 35.45 | 37.75 | — | |
| FGAHOIArchitecture=One-Stage / Transformer-Based, Backbone=Swin-Large*, Fine-tuned=false, Feature=A2023.01 | 37.18 | — | — | — | — | — | — | 30.71 | 39.11 | — | |
| Ma et al.Evaluation setting=Default setting2025.03 | 37.18 | — | — | — | — | — | — | 30.71 | 39.11 | — | |
| FGAHOIBackbone=Swin-L, Closed-set setting=true2025.12 | 37.18 | — | — | — | — | — | — | 30.71 | 39.11 | — | |
| ScriptHOI2026.05 | 37.1 | — | — | — | — | — | — | 37.8 | 36.9 | — | |
| GEN-VLKT-BDetector=HICO-DET, Backbone=ResNet50, Evaluation Setting=Known Object2023.03 | 36.78 | — | — | — | — | — | — | 32.75 | 37.99 | — | |
| AGERBackbone=ResNet-50, Closed-set setting=true2025.12 | 36.75 | — | — | — | — | — | — | 33.53 | 37.71 | — | |
| DP-HOIBackbone=ResNet-50, Closed-set setting=true2025.12 | 36.56 | — | — | — | — | — | — | 34.36 | 37.22 | — | |
| IFDetector=HICO-DET, Backbone=ResNet50, Evaluation Setting=Known Object2023.03 | 36.28 | — | — | — | — | — | — | 33.16 | 37.21 | — | |
| Zhang et al. [62]Evaluation setting=Known objects setting2025.03 | 36.08 | — | — | — | — | — | — | 31.6 | 37.47 | — | |
| GEN-VLKT-L + Category Query LearningDetector=HICO-DET, Backbone=ResNet101, Evaluation Setting=Default2023.03 | 36.03 | — | — | — | — | — | — | 33.16 | 36.89 | — | |
| CQLBackbone=ResNet-101, Closed-set setting=true2025.12 | 36.03 | — | — | — | — | — | — | 33.16 | 36.89 | — | |
| LAINEvaluation Protocol=Fully-supervised2025.05 | 36.02 | — | — | — | — | — | — | 35.7 | 36.11 | — | |
| LAINSupervision Setting=fully supervised2026.02 | 36.02 | — | — | — | — | — | — | 35.7 | 36.11 | — | |
| GEN-VLKT-M + Category Query LearningDetector=HICO-DET, Backbone=ResNet101, Evaluation Setting=Default2023.03 | 35.83 | — | — | — | — | — | — | 32.91 | 36.7 | — | |
| QAHOIArchitecture=One-Stage / Transformer-Based, Backbone=Swin-Large*, Fine-tuned=false, Feature=A2023.01 | 35.78 | — | — | — | — | — | — | 29.8 | 37.56 | — | |
| Chen et al.Evaluation setting=Default setting2025.03 | 35.78 | — | — | — | — | — | — | 29.8 | 37.56 | — | |
| LOGICHOIBackbone=ResNet-502024.08 | 35.47 | — | — | — | — | — | — | 32.03 | 36.22 | — | |
| LogicHOIEvaluation Protocol=Fully-supervised2025.05 | 35.47 | — | — | — | — | — | — | 32.03 | 36.22 | — | |
| QAHOI (R)Architecture=One-Stage / Transformer-Based, Backbone=Swin-Large*, Fine-tuned=false, Feature=A2023.01 | 35.43 | — | — | — | — | — | — | 29.22 | 37.29 | — | |
| DEFRBackbone=ViT-L/14, Fine-tuned on HICO-DET=false2021.12 | 35.36 | — | — | — | — | — | — | 38.28 | 34.49 | — | |
| GEN-VLKT-B + Category Query LearningDetector=HICO-DET, Backbone=ResNet50, Evaluation Setting=Default2023.03 | 35.36 | — | — | — | — | — | — | 32.97 | 36.07 | — | |
| CLIP4HOIBackbone=ResNet-502024.08 | 35.33 | — | — | — | — | — | — | 33.95 | 35.74 | — | |
| CLIP4HOIEvaluation Protocol=Zero-shot2025.05 | 35.33 | — | — | — | — | — | — | 33.95 | 35.74 | — | |
| CLIP4HOIBackbone=ResNet-50, Closed-set setting=true2025.12 | 35.33 | — | — | — | — | — | — | 33.95 | 35.74 | — | |
| CLIP4HOISupervision Setting=fully supervised2026.02 | 35.33 | — | — | — | — | — | — | 33.95 | 35.74 | — | |
| CLIP4HOI2026.05 | 35.3 | — | — | — | — | — | — | 32.5 | 36.1 | — | |
| BodyPartMapDetector=HICO-DET, Backbone=ResNet50, Evaluation Setting=Default2023.03 | 35.15 | — | — | — | — | — | — | 33.71 | 35.58 | — | |
| Liao et al.Evaluation setting=Default setting2025.03 | 34.95 | — | — | — | — | — | — | 31.5 | 36.08 | — | |
| GEN-VLKT-LDetector=HICO-DET, Backbone=ResNet101, Evaluation Setting=Default2023.03 | 34.95 | — | — | — | — | — | — | 31.18 | 36.08 | — | |
| HOIGenBackbone=ResNet-502024.08 | 34.84 | — | — | — | — | — | — | 34.52 | 34.94 | — | |
| GEN-VLKT2026.05 | 34.8 | — | — | — | — | — | — | 31.2 | 35.9 | — | |
| GEN-VLKT-MDetector=HICO-DET, Backbone=ResNet101, Evaluation Setting=Default2023.03 | 34.78 | — | — | — | — | — | — | 31.5 | 35.77 | — | |
| HOICLIPBackbone=ResNet-50, Training-free enhancement=true2023.03 | 34.69 | — | — | — | — | — | — | 31.12 | 35.74 | — | |
| HOICLIPExtras=L2023.11 | 34.69 | — | — | — | — | — | — | 31.12 | 35.74 | — | |
| PVICBackbone=ResNet-502024.08 | 34.69 | — | — | — | — | — | — | 32.14 | 35.45 | — | |
| HOICLIPBackbone=ResNet-502024.08 | 34.69 | — | — | — | — | — | — | 31.12 | 35.74 | — | |
| HOICLIPEvaluation Protocol=Zero-shot2025.05 | 34.69 | — | — | — | — | — | — | 31.12 | 35.74 | — | |
| HOICLIPBackbone=ResNet-50, Closed-set setting=true2025.12 | 34.69 | — | — | — | — | — | — | 31.12 | 35.74 | — | |
| HOICLIPSupervision Setting=fully supervised2026.02 | 34.69 | — | — | — | — | — | — | 31.12 | 35.74 | — | |
| HOICLIPSupervision=Fully Supervised, Detector=DETR, Vision Backbone=CLIP-B, Text Backbone=CLIP-B2026.04 | 34.69 | — | — | — | — | — | — | 31.12 | 35.74 | — | |
| HOICLIPBackbone=ResNet-50, Training-free enhancement=false2023.03 | 34.54 | — | — | — | — | — | — | 30.71 | 35.7 | — | |
| CDNDetector=HICO-DET, Backbone=ResNet50, Evaluation Setting=Known Object2023.03 | 34.53 | — | — | — | — | — | — | 29.73 | 35.96 | — | |
| DTDetector=HICO-DET, Backbone=ResNet50, Evaluation Setting=Known Object2023.03 | 34.5 | — | — | — | — | — | — | 30.13 | 35.81 | — | |
| STIPDetector=HICO-DET, Backbone=ResNet50, Evaluation Setting=Known Object2023.03 | 34.41 | — | — | — | — | — | — | 30.12 | 35.69 | — | |
| SCGDetector=HICO-DET, Backbone=ResNet50-FPN, Evaluation Setting=Known Object2023.03 | 34.37 | — | — | — | — | — | — | 27.18 | 36.52 | — | |
| MSTRDetector=HICO-DET, Backbone=ResNet50, Evaluation Setting=Known Object2023.03 | 34.02 | — | — | — | — | — | — | 28.83 | 35.57 | — | |
| ADA-CMBackbone=ResNet-502024.08 | 33.8 | — | — | — | — | — | — | 31.72 | 34.42 | — | |
| ADA-CMEvaluation Protocol=Fully-supervised2025.05 | 33.8 | — | — | — | — | — | — | 31.72 | 34.42 | — | |
| ADA-CMSupervision Setting=fully supervised2026.02 | 33.8 | — | — | — | — | — | — | 31.72 | 34.42 | — | |
| ADA-CMSupervision=Fully Supervised, Detector=DETR, Vision Backbone=CLIP-B, Text Backbone=CLIP-B2026.04 | 33.8 | — | — | — | — | — | — | 31.72 | 34.42 | — | |
| GEN-VLKTBackbone=ResNet-502023.03 | 33.75 | — | — | — | — | — | — | 29.25 | 35.1 | — |