Research map / Detection and segmentation / Open-set and few-shot detection
Discovering Human-Object Interaction Concepts via Self-Compositional Learning
Zhi Hou, Baosheng Yu, Dacheng Tao
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 self-training with pseudo labels to online concept discovery (ablation) | SCL · HICO-DET unknown concepts | AP: 22.25 -> 33.58 | Improved |
| add self-training with pseudo labels to online concept discovery (ablation) | SCL · V-COCO unknown concepts | AP: 24.89 -> 28.77 | Improved |
| add self-training on concept confidence to Qpic | Qpic* · HICO-DET unknown concepts | AP: 27.53 -> 28.44 | Improved |
| add the spatial branch (ablation) | SCL · HICO-DET unknown concepts | AP: 33.26 -> 32.56 | Hurt |
| add verb auxiliary loss (ablation) | SCL · HICO-DET unknown concepts | AP: 29.56 -> 32.56 | Improved |
| extract verb features from the union box instead of the human box (ablation) | SCL · HICO-DET unknown concepts | AP: 28.30 -> 32.56 | Improved |
| increase the self-training loss weight lambda3 from 0.5 to 2 (ablation) | SCL · V-COCO unknown concepts | AP: 29.52 -> 31.33 | Improved |
| lower temperature T from 1 to 0.25 (ablation) | SCL · HICO-DET unknown concepts (1M iters, no spatial branch) | AP: 30.36 -> 33.66 | Improved |
| normalize the confidence matrix for pseudo labels (ablation) | SCL · HICO-DET unknown concepts | AP: 32.30 -> 32.56 | No clear effect |
| use better refined concept annotations for evaluation | SCL · HICO-DET unknown concepts | AP: 33.26 -> 33.58 | No clear effect |
| add concept discovery (top 120 concepts) to HOI detection, rare first | SCL · HICO-DET zero-shot unknown, rare first | mAP: 1.68 -> 2.26 | Improved |
| add concept discovery (top 120 concepts) to HOI detection, non-rare first | SCL · HICO-DET zero-shot unknown, non-rare first | mAP: 5.86 -> 7.05 | Improved |
Closest papers by meaning
- Detecting Human-Object Interaction via Fabricated Compositional Learning (CVPR 2021 · similarity 0.93)
- Detecting Any Human-Object Interaction Relationship: Universal HOI Detector with Spatial Prompt Learning on Foundation Models (NeurIPS 2023 · similarity 0.91)
- Exploring the Potential of Large Foundation Models for Open-Vocabulary HOI Detection (CVPR 2024 · similarity 0.89)
- Open-World Human-Object Interaction Detection via Multi-modal Prompts (CVPR 2024 · similarity 0.89)
- Human-Object Interaction Detection Collaborated with Large Relation-driven Diffusion Models (NeurIPS 2024 · similarity 0.88)
- HOIMamba: Efficient Mamba-based Disentangled Progressive Learning for HOI Detection (AAAI 2025 · 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.
