Research map / Detection and segmentation / Weakly supervised detection and segmentation
Weakly-supervised HOI Detection via Prior-guided Bi-level Representation Learning
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 CLIP-based HOI recognition to the CLIP RN50 pairwise baseline (ablation) | Weakly-supervised HOI detection · HICO-DET | Full mAP: 19.52 to 20.31 | Helped |
| Freeze the HOI knowledge bank instead of jointly learning it (ablation) | HOI recognition model · HICO-DET | Full mAP: 20.31 to 20.09 | Hurt |
| Add KTN using the HOI knowledge bank (ablation) | HOI recognition model · HICO-DET | Full mAP: +0.55 Full mAP | Helped |
| Fuse global HOI scores with pairwise interaction scores (ablation) | HOI recognition and KTN model · HICO-DET | Full mAP: 20.86 to 22.40 | Helped |
| Add self-taught relatedness classification (ablation) | HOI recognition, KTN, and score-fusion model · HICO-DET | Full mAP: 22.40 to 22.89 | Helped |
| Add SRC to the baseline without HOI knowledge (ablation) | Weakly-supervised HOI detection · HICO-DET | Full mAP: 19.52 to 19.88 | Helped |
| Add KTN to the model with SRC (ablation) | HOI recognition and SRC model · HICO-DET | Full mAP: 21.53 to 20.75 | Helped |
| Replace CLIP visual initialization with COCO detection-trained RN50-FPN (ablation) | Weakly-supervised HOI detection · HICO-DET | Full mAP: 19.44 vs 22.89 | Hurt |
| Replace the CLIP text knowledge bank with random initialization (ablation) | Full model with COCO RN50-FPN · HICO-DET | Rare mAP: 16.20 to 15.57 | Hurt |
| Add CLIP inference scores to the COCO RN50-FPN baseline (ablation) | Weakly-supervised HOI detection · HICO-DET | Full mAP: 19.44 to 20.56 | Helped |
| Apply HOI-prompt classifier initialization and CLIP visual regularization (ablation) | COCO RN50-FPN baseline · HICO-DET | Full mAP: 19.44 to 19.39 | Hurt |
| Use union-region visual features instead of KTN knowledge-bank attention (ablation) | HOI recognition and SRC model · HICO-DET | Full mAP: 20.75 to 20.69 | Hurt |
Closest papers by meaning
- HOICLIP: Efficient Knowledge Transfer for HOI Detection With Vision-Language Models (CVPR 2023 · similarity 0.92)
- Learning to Detect Human-Object Interactions With Knowledge (CVPR 2019 · similarity 0.92)
- LINK: Learning Instance-level Knowledge from Vision-Language Models for Human-Object Interaction Detection (ICLR 2026 · similarity 0.92)
- Detecting Human-Object Interaction via Fabricated Compositional Learning (CVPR 2021 · similarity 0.91)
- Exploring Pose-Aware Human-Object Interaction via Hybrid Learning (CVPR 2024 · similarity 0.90)
- HORP: Human-Object Relation Priors Guided HOI Detection (CVPR 2025 · 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.