Research map / Detection and segmentation / Open-set and few-shot detection
Chairs Can Be Stood On: Overcoming Object Bias in Human-Object Interaction Detection
Guangzhi Wang, Yangyang Guo, Yongkang Wong, Mohan Kankanhalli
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 the object-debiasing method (ODM) | SCG · HICO-DET KO pretrained detector | Full mAP: 24.53 -> 25.54 | Improved |
| add the object-debiasing method | SCG · HICO-DET KO pretrained detector | Rare mAP: 20.00 -> 21.93 | Improved |
| add the object-debiasing method | SCG · HICO-DET KO fine-tuned detector | Full mAP: 33.74 -> 34.52 | Improved |
| add the object-debiasing method | SCG · HICO-DET KO fine-tuned detector | Rare mAP: 26.41 -> 27.34 | Improved |
| set lambda to 0.4 rather than 0.2 (ablation) | SCG + Ours · HICO-DET | Full mAP: 21.32 -> 21.50 | Improved |
| raise lambda from 0.4 to 0.8 (ablation) | SCG + Ours · HICO-DET | Full mAP: 21.50 -> 21.17 | Hurt |
| add the method with batch of 4 GPUs x 1 image | SCG · HICO-DET | Full mAP: 20.75 -> 21.16 | Improved |
| add the method with batch of 4 GPUs x 2 image | SCG · HICO-DET | Full mAP: 20.99 -> 21.50 | Improved |
| use batch of 4 GPUs x 2 images instead of 1 image (ablation) | SCG baseline · HICO-DET | Full mAP: 20.75 -> 20.99 | Improved |
| scale learning rate linearly with batch size (ablation) | SCG baseline · HICO-DET | Full mAP: 19.94 -> 20.99 | Improved |
| add the method overhead | SCG · Efficiency | train time per image: 428.16ms -> 440.72ms | No clear effect |
| apply re-weighting, adversarial or domain-independent debiasing instead of ours | SCG · HICO-DET | Full mAP | Hurt |
Closest papers by meaning
- Interaction Confidence Attention for Human–Object Interaction Detection (IJCV 2025 · similarity 0.90)
- No-Frills Human-Object Interaction Detection: Factorization, Layout Encodings, and Training Techniques (ICCV 2019 · similarity 0.86)
- Learning Human-Object Interaction Detection Using Interaction Points (CVPR 2020 · similarity 0.85)
- Disentangled Pre-training for Human-Object Interaction Detection (CVPR 2024 · similarity 0.85)
- Orchestrating the Symphony of Prompt Distribution Learning for Human-Object Interaction Detection (AAAI 2025 · similarity 0.82)
- Exploring Pose-Aware Human-Object Interaction via Hybrid Learning (CVPR 2024 · similarity 0.82)
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.
