Research map / Detection and segmentation / Open-set and few-shot detection
Relational Context Learning for Human-Object Interaction Detection
Sanghyun Kim, Deunsol Jung, Minsu Cho
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 |
|---|---|---|---|
| Use three branches with multiplex relational context exchange | MUREN · HICO-DET Default and Known Object; V-COCO | mAP / AP_role | Improved |
| Add ternary relation context to the no-context baseline (ablation) | MUREN · V-COCO | AP_role: +4.55%p scenario 1; +4.22%p scenario 2 | Improved |
| Add unary context to ternary context (ablation) | MUREN · V-COCO | AP_role: 67.07 -> 68.12 (scenario 1); 69.36 -> 70.31 (scenario 2) | Improved |
| Add pairwise context to ternary context (ablation) | MUREN · V-COCO | AP_role: 67.07 -> 67.67 (scenario 1); 69.36 -> 70.02 (scenario 2) | Improved |
| Use all ternary, unary, and pairwise relation contexts (ablation) | MUREN · V-COCO | AP_role: +6.23%p scenario 1; +5.86%p scenario 2 | Improved |
| Propagate multiplex context to the human branch (ablation) | MUREN · V-COCO | AP_role: 62.52 -> 64.44 (scenario 1); 65.14 -> 66.62 (scenario 2) | Improved |
| Propagate multiplex context to the object branch (ablation) | MUREN · V-COCO | AP_role: 62.52 -> 63.66 (scenario 1); 65.14 -> 66.00 (scenario 2) | Improved |
| Propagate multiplex context to the interaction branch (ablation) | MUREN · V-COCO | AP_role: +3.19%p scenario 1; +2.77%p scenario 2 | Improved |
| Propagate multiplex context to all three branches (ablation) | MUREN · V-COCO | AP_role: 62.52 -> 68.75 (scenario 1); 65.14 -> 71.00 (scenario 2) | Improved |
| Add task-conditioned context transformation (ablation) | MUREN attentive fusion · V-COCO | AP_role: 66.50 -> 66.95 (scenario 1); 68.96 -> 69.23 (scenario 2) | Improved |
| Add channel attention (ablation) | MUREN attentive fusion · V-COCO | AP_role: 66.50 -> 67.10 (scenario 1); 68.96 -> 69.49 (scenario 2) | Improved |
| Use both task conditioning and channel attention (ablation) | MUREN attentive fusion · V-COCO | AP_role: 66.50 -> 68.75 (scenario 1); 68.96 -> 71.00 (scenario 2) | Improved |
Closest papers by meaning
- Cascaded Human-Object Interaction Recognition (CVPR 2020 · similarity 0.91)
- Detecting Any Human-Object Interaction Relationship: Universal HOI Detector with Spatial Prompt Learning on Foundation Models (NeurIPS 2023 · similarity 0.91)
- Toward Open-Set Human Object Interaction Detection (AAAI 2024 · similarity 0.88)
- Learning Human-Object Interaction as Groups (NeurIPS 2025 · similarity 0.88)
- HOIMamba: Efficient Mamba-based Disentangled Progressive Learning for HOI Detection (AAAI 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.
