Research map / Detection and segmentation / Open-set and few-shot detection
No More Sibling Rivalry: Debiasing Human-Object Interaction Detection
Bin Yang, Yulin Zhang, Hong-Yu Zhou, Sibei Yang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add human-object region mask attention (ablation) | GEN-VLKT-s · HICO-DET Default Full | mAP: 33.75 -> 35.37 | Improved |
| add contrastive loss on sibling triplets (ablation) | GEN-VLKT-s + HOR mask · HICO-DET Default Full | mAP: 35.37 -> 36.82 | Improved |
| add calibration loss (ablation) | GEN-VLKT-s + HOR mask + contrastive · HICO-DET Default Full | mAP: 36.82 -> 38.03 | Improved |
| add merge loss (superclass hierarchical) (ablation) | GEN-VLKT-s + C2C · HICO-DET Default Full | mAP: 38.03 -> 39.88 | Improved |
| add split loss (ablation) | GEN-VLKT-s + C2C + merge · HICO-DET Default Full | mAP: 39.88 -> 42.93 | Improved |
| add C2C and M2S objectives (ablation) | HOICLIP · HICO-DET Default Full | mAP: 34.54 -> 43.11 | Improved |
| add HOR mask on top of C2C and M2S (ablation) | Ours · HICO-DET Default Full | mAP: 42.01 -> 42.93 | Improved |
| vary split range k2 (5 vs 10) (ablation) | Ours (k1=2) · HICO-DET Default Full | mAP: 41.20 -> 42.93 | Hurt |
| vary split range k2 (15 vs 10) (ablation) | Ours (k1=2) · HICO-DET Default Full | mAP: 42.72 -> 42.93 | Hurt |
| vary cluster count (3x3 vs 5x5) (ablation) | Ours · HICO-DET Default Full | mAP: 41.88 -> 42.93 | Hurt |
| vary cluster count (7x7 vs 5x5) (ablation) | Ours · HICO-DET Default Full | mAP: 42.15 -> 42.93 | Hurt |
| vary k1 (1 vs 2) (ablation) | Ours (k2=10) · V-COCO AP S1 role | AP: 69.0 -> 69.8 | Hurt |
Closest papers by meaning
- Human-Object Interaction Detection via Disentangled Transformer (CVPR 2022 · similarity 0.91)
- Learning Human-Object Interaction Detection Using Interaction Points (CVPR 2020 · similarity 0.89)
- Exploring Self- and Cross-Triplet Correlations for Human-Object Interaction Detection (AAAI 2024 · similarity 0.89)
- Exploring Pose-Aware Human-Object Interaction via Hybrid Learning (CVPR 2024 · similarity 0.89)
- Learning Human-Object Interaction as Groups (NeurIPS 2025 · similarity 0.88)
- HORP: Human-Object Relation Priors Guided HOI Detection (CVPR 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
