Research map / Detection and segmentation / Open-set and few-shot detection
UniHOI: Unified Human-Object Interaction Understanding via Unified Token Space
Panqi Yang, Haodong Jing, Nanning Zheng, Yongqiang Ma
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 modality-type embedding to unified token space (ablation) | UniHOI · HICO-DET Default Full | mAP: 47.62 -> 48.16 | Improved |
| unify vocabulary and embeddings instead of separate embeddings (ablation) | UniHOI · HICO-DET Default Full | mAP: 47.03 -> 48.16 | Improved |
| unified token space vs separate embeddings on generation (ablation) | UniHOI · HOI generation | HOI Score: 0.44 -> 0.64 | Improved |
| add modality-type embedding on generation (ablation) | UniHOI · HOI generation | HOI Score: 0.57 -> 0.64 | Improved |
| add interaction-aware attention (IAA) (ablation) | UniHOI · HOI generation | Interaction Accuracy: 0.39 -> 0.54 | Improved |
| add IAA to detection (ablation) | UniHOI · HICO-DET Default Rare | mAP: 49.03 -> 50.74 | Improved |
| add weak and unsupervised data (10/25/65 vs 100% strong) (ablation) | UniHOI · HICO-DET Default Full | mAP: 45.62 -> 48.16 | Improved |
| add weak and unsupervised data on generation (ablation) | UniHOI · HOI generation | HOI Score: 0.22 -> 0.64 | Improved |
| UniHOI vs HOI-IDiff on V-COCO scenario 1 | UniHOI · V-COCO Scenario 1 | AP role: 73.42 -> 72.91 | Hurt |
| UniHOI vs SDv3.5 on FID | UniHOI · HOI generation | FID: 17.7 -> 18.2 | Hurt |
| UniHOI vs best baseline on interaction metrics | UniHOI · HOI generation | HOI Score: 0.56 -> 0.64 | Improved |
Closest papers by meaning
- Detecting Any Human-Object Interaction Relationship: Universal HOI Detector with Spatial Prompt Learning on Foundation Models (NeurIPS 2023 · similarity 0.92)
- Exploring Pose-Aware Human-Object Interaction via Hybrid Learning (CVPR 2024 · similarity 0.89)
- Exploring the Potential of Large Foundation Models for Open-Vocabulary HOI Detection (CVPR 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)
- Open-World Human-Object Interaction Detection via Multi-modal Prompts (CVPR 2024 · 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.
