Research map / Detection and segmentation / Open-set and few-shot detection
Discovering Human Interactions With Novel Objects via Zero-Shot Learning
Suchen Wang, Kim-Hui Yap, Junsong Yuan, Yap-Peng Tan
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 |
|---|---|---|---|
| use HO-RPN to score object proposals by their interactiveness with humans (ablation) | HOI detection on V-COCO · V-COCO (interacting object detection) | AP@IoU=0.5: 29.4 -> 42.7 | Improved |
| use HO-RPN to score object proposals by their interactiveness with humans (ablation) | HOI detection on HICO-DET · HICO-DET (interacting object detection) | AP@IoU=0.5: 33.6 -> 45.7 | Improved |
| use HO-RPN rather than RPN w/ FPN for object proposals | V-COCO val · V-COCO val (all objects, top 100 proposals) | Recall@100: 87.1 -> 89.8 | Improved |
| use 12 human boxes rather than 8 (ablation) | HO-RPN · V-COCO (interacting object detection) | AP@IoU=0.5: 42.7 -> 42.8 | No clear effect |
| use 12 human boxes rather than 8 (ablation) | HO-RPN · HICO-DET (interacting object detection) | AP@IoU=0.5: 45.7 -> 45.9 | No clear effect |
| use GoogleNews word2vec embeddings rather than GloVe (ablation) | ZS-HOID · V-COCO (novel object detection) | AP@IoU=0.5: 10.4 -> 11.5 | Improved |
| use GoogleNews word2vec embeddings rather than FastText (ablation) | ZS-HOID · V-COCO (novel object detection) | AP@IoU=0.5: 10.2 -> 11.5 | Improved |
| use HO-RPN with interactiveness scores rather than zero-shot object detection proposals | ZS-HOID · V-COCO (novel object detection) | AP@IoU=0.5: 3.3 -> 11.5 | Improved |
| use a seen/unseen split with related seen categories (ablation) | zero-shot object recognition · V-COCO and HICO-DET seen/unseen splits | AP@IoU=0.5 | Improved |
| add human-visual interactiveness scoring (ablation) | ZS-HOID · HICO-DET test (unseen) | mAP (%): 1.45 -> 3.02 | Improved |
Closest papers by meaning
- Detecting and Recognizing Human-Object Interactions (CVPR 2018 · similarity 0.90)
- Interaction Compass: Multi-Label Zero-Shot Learning of Human-Object Interactions via Spatial Relations (ICCV 2021 · similarity 0.90)
- Visual Diversity and Region-aware Prompt Learning for Zero-shot HOI Detection (NeurIPS 2025 · similarity 0.89)
- Locality-Aware Zero-Shot Human-Object Interaction Detection (CVPR 2025 · similarity 0.89)
- Toward Open-Set Human Object Interaction Detection (AAAI 2024 · similarity 0.88)
- Human-Object Interaction Detection Collaborated with Large Relation-driven Diffusion Models (NeurIPS 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.
