Research map / Detection and segmentation / Open-set and few-shot detection
Diagnosing Human-Object Interaction Detectors
Fangrui Zhu, Yiming Xie, Weidi Xie, Huaizu Jiang
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 |
|---|---|---|---|
| apply NMS to remove duplicate pairs (ablation) | one-stage CDN · HICO-DET | Pair Precision: 2.5 -> 5.5 | Improved |
| apply NMS to remove duplicate pairs (ablation) | one-stage CDN · HICO-DET | Pair Recall: 79.4 -> 72.0 | Hurt |
| apply NMS to remove duplicate pairs (ablation) | one-stage CDN · HICO-DET | HOI mAP: 34.4 -> 34.4 | No clear effect |
| lower object detection threshold to get more pairs (ablation) | two-stage SCG · HICO-DET | Pair Recall: 84.0 -> 86.9 | Improved |
| lower object detection threshold to get more pairs (ablation) | two-stage SCG · HICO-DET | Pair Precision: 3.7 -> 2.2 | Hurt |
| lower object detection threshold to get more pairs (ablation) | two-stage SCG · HICO-DET | HOI mAP: 33.3 -> 33.3 | No clear effect |
| stronger backbone SwinT to SwinL | RLIPv2 · HICO-DET | HOI mAP: 41.6 -> 49.1 | Improved |
| stronger backbone ResNet50 to ResNet101 | UPT · HICO-DET | HOI mAP: 34.6 -> 35.5 | Improved |
| stronger backbone ResNet50 to ResNet101 | GEN-VLKT · HICO-DET | HOI mAP: 35.8 -> 36.5 | Improved |
| stronger backbone SwinT to SwinL | RLIPv2 · HICO-DET | Interaction mAP: 56.7 -> 61.1 | Improved |
| large-scale relational pre-training (RLIPv2) vs other methods | RLIPv2 · HICO-DET | Interaction mAP: 46.5 -> 56.7 | Improved |
| evaluate on rare instead of non-rare categories | RLIPv2 SwinL · HICO-DET | Interaction mAP: 54.4 -> 21.7 | Hurt |
Closest papers by meaning
- Learning Human-Object Interaction Detection Using Interaction Points (CVPR 2020 · similarity 0.92)
- Mining the Benefits of Two-stage and One-stage HOI Detection (NeurIPS 2021 · similarity 0.92)
- Detecting Human-Object Interactions via Functional Generalization (AAAI 2020 · similarity 0.92)
- HORP: Human-Object Relation Priors Guided HOI Detection (CVPR 2025 · similarity 0.90)
- Exploring Pose-Aware Human-Object Interaction via Hybrid Learning (CVPR 2024 · similarity 0.90)
- Exploring the Potential of Large Foundation Models for Open-Vocabulary HOI Detection (CVPR 2024 · similarity 0.90)
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.
