Research map / Detection and segmentation / Open-set and few-shot detection
CLIP4HOI: Towards Adapting CLIP for Practical Zero-Shot HOI Detection
Yunyao Mao, Jiajun Deng, Wengang Zhou, Li Li, Yao Fang, Houqiang Li
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 HOI decoder aggregating CLIP visual features (ablation) | CLIP4HOI · HICO-DET RF-UC | unseen mAP: 23.80 -> 28.47 | Improved |
| add HO interactor (pairwise interaction and spatial injection) (ablation) | CLIP4HOI · HICO-DET RF-UC | full mAP: 32.98 -> 34.08 | Improved |
| add CLIP class token as decoder input (ablation) | CLIP4HOI · HICO-DET RF-UC | full mAP: 33.50 -> 34.08 | Improved |
| aggregate mid-level CLIP features instead of final features in decoder (ablation) | CLIP4HOI · HICO-DET RF-UC | full mAP: 34.08 -> 32.83 | Hurt |
| add global HOI score (ablation) | CLIP4HOI · HICO-DET RF-UC | unseen mAP: 25.48 -> 28.47 | Improved |
| add global HOI score (ablation) | CLIP4HOI · HICO-DET RF-UC | seen mAP: 36.06 -> 35.48 | Hurt |
| add DETR object class priors (ablation) | CLIP4HOI · HICO-DET RF-UC | unseen mAP: 24.68 -> 28.47 | Improved |
| use learnable prompt instead of manual prompt (ablation) | CLIP4HOI · HICO-DET RF-UC | unseen mAP: 23.92 -> 28.47 | Improved |
| increase HOI decoder layers beyond 6 (ablation) | CLIP4HOI · HICO-DET RF-UC | full mAP: 34.08 -> 33.94 | No clear effect |
| set pairwise loss weight beta to 1.0 vs 5.0 (ablation) | CLIP4HOI · HICO-DET RF-UC | unseen mAP: 26.41 -> 28.47 | Improved |
| use two-stage disentangled pipeline instead of one-stage EoID under large position distribution discrepancy | HICO-DET NF-UC unseen · HICO-DET NF-UC | unseen mAP: 18.92 -> 28.23 | Improved |
Later papers that compared with this method
2 later accepted papers report CLIP4HOI as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- HOLa: Zero-Shot HOI Detection with Low-Rank Decomposed VLM Feature Adaptation (ICCV 2025 · better on 3 of 3 shared results)
- Open-Vocabulary HOI Detection with Interaction-aware Prompt and Concept Calibration (ICCV 2025 · better on 0 of 2 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Locality-Aware Zero-Shot Human-Object Interaction Detection (CVPR 2025 · similarity 0.93)
- Exploring the Potential of Large Foundation Models for Open-Vocabulary HOI Detection (CVPR 2024 · similarity 0.89)
- Visual Diversity and Region-aware Prompt Learning for Zero-shot HOI Detection (NeurIPS 2025 · similarity 0.88)
- HORP: Human-Object Relation Priors Guided HOI Detection (CVPR 2025 · similarity 0.88)
- EZ-HOI: VLM Adaptation via Guided Prompt Learning for Zero-Shot HOI Detection (NeurIPS 2024 · similarity 0.88)
- Open-World Human-Object Interaction Detection via Multi-modal Prompts (CVPR 2024 · similarity 0.87)
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.
