Research map / Detection and segmentation / Open-set and few-shot detection
Detecting Any Human-Object Interaction Relationship: Universal HOI Detector with Spatial Prompt Learning on Foundation Models
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 BLIP2 foundation model with HO Prompt-guided Decoder to GEN-VLKT-s (ablation) | GEN-VLKT-s · V-COCO | AP_role #1: 62.91 -> 65.58 | Helped |
| add BLIP2 naively via learnable Q-Former queries (no HOPD) (ablation) | GEN-VLKT-s · V-COCO | AP_role #1: 62.41 -> 62.91 | No clear effect |
| add HOPD spatial prompting (ablation) | GEN-VLKT-s + BLIP2 · HICO-DET UV Unseen | mAP: 21.57 -> 26.05 | Helped |
| add GPT knowledge retrieval descriptive texts (ablation) | UniHOI-s · V-COCO | AP_role #1: 65.58 -> 66.74 | Helped |
| add GPT knowledge retrieval descriptive texts (ablation) | UniHOI-s · HICO-DET UV Unseen | mAP: 26.05 -> 27.41 | Helped |
| use BLIP2 instead of CLIP as foundation model (ablation) | UniHOI-s · V-COCO | AP_role #1: 62.92 -> 65.58 | Helped |
| use larger BLIP-2 (ViT-G OPT6.7B) (ablation) | UniHOI-s · V-COCO | AP_role #1: 65.58 -> 65.92 | Helped |
| use FlanT5-XL LLM variant instead of OPT2.7B (ablation) | UniHOI-s · V-COCO | AP_role #1: 65.58 -> 65.37 | No clear effect |
| replace prior best with UniHOI (BLIP2 + spatial prompts) | GEN-VLKT-s · HICO-DET Default Full | mAP: 33.75 -> 40.06 | Helped |
Closest papers by meaning
- HORP: Human-Object Relation Priors Guided HOI Detection (CVPR 2025 · similarity 0.93)
- Exploring the Potential of Large Foundation Models for Open-Vocabulary HOI Detection (CVPR 2024 · similarity 0.92)
- Open-World Human-Object Interaction Detection via Multi-modal Prompts (CVPR 2024 · similarity 0.92)
- Exploring Pose-Aware Human-Object Interaction via Hybrid Learning (CVPR 2024 · similarity 0.92)
- Toward Open-Set Human Object Interaction Detection (AAAI 2024 · similarity 0.91)
- Learning Human-Object Interaction as Groups (NeurIPS 2025 · similarity 0.91)
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.