HOI Detection on V-COCO v1 (test)
72.1AP Role (Scenario 1)RLIPv2-ParSeDA
Evaluation results
| Method | Links | ||
|---|---|---|---|
| RLIPv2-ParSeDABackbone=Swin-L, Extra Relations=VG+COCO+O3652023.08 | 72.1 | 74.1 | |
| RLIPv2-ParSeDABackbone=Swin-T, Extra Relations=VG+COCO+O3652023.08 | 68.8 | 70.8 | |
| MP-HOI-SBackbone=ResNet-50, Training Data Scale=100% Training Data2024.06 | 66.2 | 67.6 | |
| RLIPv2-ParSeDABackbone=R50, Extra Relations=VG+COCO+O3652023.08 | 65.9 | 68 | |
| UniVRDBackbone=ViT-H/14, Pre-training=LiT + O365 + COCO + HICO2023.08 | 65.8 | 66.9 | |
| RLIPv2-ParSeDABackbone=R50, Extra Relations=VG+COCO2023.08 | 64.5 | 66.7 | |
| OCNBackbone=R502023.08 | 64.2 | 66.3 | |
| RLIPv2-ParSeDABackbone=R50, Extra Relations=VG2023.08 | 63 | 65.1 | |
| GEN-VLKTBackbone=R502023.08 | 62.4 | 64.5 | |
| GEN-VLKTBackbone=ResNet-50, Training Data Scale=100%2024.06 | 62.4 | 64.5 | |
| RLIPv1-ParSeBackbone=R50, Extra Relations=VG2023.08 | 61.9 | 64.2 | |
| CDNBackbone=R502023.08 | 61.7 | 63.8 | |
| RLIPv1-ParSeDBackbone=R50, Extra Relations=VG2023.08 | 61.7 | 63.8 | |
| CDNBackbone=ResNet-50, Training Data Scale=100%2024.06 | 61.7 | 63.8 | |
| QPICBackbone=R502023.08 | 58.8 | 61 | |
| QPICBackbone=ResNet-50, Training Data Scale=100%2024.06 | 58.8 | 61 | |
| MP-HOI-SBackbone=ResNet-50, Training Data Scale=10% Training Data2024.06 | 57.7 | 60.2 | |
| HOTRBackbone=R502023.08 | 55.2 | 64.4 | |
| HOTRBackbone=ResNet-50, Training Data Scale=100%2024.06 | 55.2 | 64.4 | |
| SCGBackbone=ResNet-50, Training Data Scale=100%2024.06 | 54.2 | 60.9 | |
| IDNBackbone=ResNet-50, Training Data Scale=100%2024.06 | 53.3 | 60.3 | |
| FCLBackbone=ResNet-50, Training Data Scale=100%2024.06 | 52.4 | — | |
| UnionDetBackbone=R50-FPN2023.08 | 47.5 | 56.2 | |
| InteractNetBackbone=R50-FPN2023.08 | 40 | — | |
| MP-HOI-SBackbone=ResNet-50, Zero-shot=true2024.06 | 37.5 | 44.2 |