Research map / Vision-language and multimodal / CLIP and zero-shot
Disentangling and Reusing Interaction Cues for Zero-Shot HOI Detection
Jie Gu, He-Yang Xu, Hongxiang Gao, Chengyu Liu
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 CORD module (ablation) | baseline CLIP two-stage HOI · HICO-DET UV | Unseen mAP: 24.88 -> 28.58 | Improved |
| add SLF module (ablation) | baseline CLIP two-stage HOI · HICO-DET UV | Unseen mAP: 24.88 -> 27.19 | Improved |
| add SLF on top of CORD (ablation) | DRIC · HICO-DET UV | Unseen mAP: 28.58 -> 29.49 | Improved |
| use orthogonal projection instead of concatenation (ablation) | CORD · HICO-DET UV | Unseen mAP: 28.41 -> 29.49 | Improved |
| use orthogonal projection instead of subtraction (ablation) | CORD · HICO-DET UV | Unseen mAP: 28.05 -> 29.49 | Improved |
| use orthogonal projection instead of cross-attention (ablation) | CORD · HICO-DET UV | Unseen mAP: 28.86 -> 29.49 | Improved |
| fuse all layers L1-L12 instead of final layer only (ablation) | SLF · HICO-DET UV | Unseen mAP: 28.58 -> 29.49 | Improved |
| fuse L6-L12 instead of L1-L12 (ablation) | SLF · HICO-DET UV | Unseen mAP: 29.03 -> 29.49 | Hurt |
| fuse L1-L6 + L12 instead of L1-L12 (ablation) | SLF · HICO-DET UV | Unseen mAP: 28.85 -> 29.49 | Hurt |
| two-stage training instead of one-stage joint (ablation) | DRIC · HICO-DET UV | Unseen mAP: 28.78 -> 29.49 | Improved |
| two-stage training instead of one-stage joint (ablation) | DRIC · HICO-DET UV | Full mAP: 33.18 -> 33.95 | Improved |
| replace background with task-irrelevant scenes (drop rate) | DRIC vs baseline · HICO-DET UV | Full mAP drop rate: 12.26% -> 2.18% | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- EZ-HOI — EZ-HOI: VLM Adaptation via Guided Prompt Learning for Zero-Shot HOI Detection (NeurIPS 2024) · better on 4 of 4 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
- End-to-End Zero-Shot HOI Detection via Vision and Language Knowledge Distillation (AAAI 2023 · similarity 0.93)
- Locality-Aware Zero-Shot Human-Object Interaction Detection (CVPR 2025 · similarity 0.93)
- Visual Diversity and Region-aware Prompt Learning for Zero-shot HOI Detection (NeurIPS 2025 · similarity 0.90)
- Exploring the Potential of Large Foundation Models for Open-Vocabulary HOI Detection (CVPR 2024 · similarity 0.90)
- Toward Open-Set Human Object Interaction Detection (AAAI 2024 · similarity 0.90)
- EZ-HOI: VLM Adaptation via Guided Prompt Learning for Zero-Shot HOI Detection (NeurIPS 2024 · similarity 0.89)
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.
