Research map / Vision-language and multimodal / CLIP and zero-shot
HOICLIP: Efficient Knowledge Transfer for HOI Detection With Vision-Language 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 |
|---|---|---|---|
| Use CLIP visual features in the interaction decoder (ablation) | HOICLIP Base · HICO-DET, Rare categories | mAP: 26.68 -> 28.74 | Helped |
| Add knowledge integration cross-attention over detector and CLIP features (ablation) | HOICLIP +CLIP · HICO-DET, Full | mAP: 32.72 -> 34.13 | Helped |
| Add a verb adapter and visual-semantic-arithmetic verb classifier (ablation) | HOICLIP +integration · HICO-DET, Full | mAP: 34.13 -> 34.54 | Helped |
| Add training-free CLIP HOI enhancement (ablation) | HOICLIP +verb · HICO-DET, Full | mAP: 34.54 -> 34.69 | Helped |
| Use visual semantic arithmetic rather than sentence-based verb representations (ablation) | HOICLIP verb classifier · HICO-DET, Full | mAP: 33.38 -> 34.54 | Helped |
| Use visual semantic arithmetic rather than averaging HOI representations (ablation) | HOICLIP verb classifier · HICO-DET, Full | mAP: 33.09 -> 34.54 | Helped |
| Set verb-logit weight alpha to 0.50 instead of 0.00 (ablation) | HOICLIP · HICO-DET validation, Full | mAP: 31.16 -> 31.65 | Helped |
| Enable training-free enhancement with Top K 10 rather than Top K 0 (ablation) | HOICLIP · HICO-DET validation, Full | mAP: 31.65 -> 32.26 | Helped |
| Replace CLIP visual encoder with ImageNet-pretrained ViT (ablation) | HOICLIP · HICO-DET, 100% training data | mAP: 34.69 -> 33.03 | Hurt |
| Use HOICLIP instead of GEN-VLKT | HOI detection · HICO-DET, 5% training data, All Categories | mAP: 13.31 -> 21.94 | Helped |
| Use HOICLIP instead of GEN-VLKT | HOI detection · HICO-DET, NF-UC Full | mAP: 23.71 -> 27.75 | Helped |
| Use HOICLIP instead of GEN-VLKT | HOI detection · HICO-DET, default Rare categories | mAP: 29.25 -> 31.12 | Helped |
Closest papers by meaning
- LINK: Learning Instance-level Knowledge from Vision-Language Models for Human-Object Interaction Detection (ICLR 2026 · similarity 0.93)
- Exploring the Potential of Large Foundation Models for Open-Vocabulary HOI Detection (CVPR 2024 · similarity 0.91)
- Open-World Human-Object Interaction Detection via Multi-modal Prompts (CVPR 2024 · similarity 0.90)
- Discovering Syntactic Interaction Clues for Human-Object Interaction Detection (CVPR 2024 · similarity 0.90)
- Locality-Aware Zero-Shot Human-Object Interaction Detection (CVPR 2025 · similarity 0.90)
- HORP: Human-Object Relation Priors Guided HOI Detection (CVPR 2025 · similarity 0.90)
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.