Research map / Detection and segmentation / Open-set and few-shot detection
Large-Scale Unsupervised Object Discovery
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 LOD (eigenvector-personalized PageRank) instead of Vo's combinatorial optimization | UOD with VGG16 · COCO C20K | AP50: 5.18 -> 6.63 | Helped |
| Use LOD instead of Vo | UOD with VGG16 · COCO C20K | CorLoc: 48.5 -> 48.5 | No clear effect |
| Add eigenvector-based personalization to PageRank (LOD vs PageRank only) (ablation) | PageRank, Vo proposals, supervised · COCO C20K | CorLoc: 47.5 -> 48.5 | Helped |
| Add PageRank on top of eigenvector ranking (LOD vs eigenvector only) (ablation) | Quadratic ranking, Vo proposals, supervised · COCO C20K | AP50: 6.21 -> 6.63 | Helped |
| Use all region proposals instead of 50 (ablation) | LOD · COCO C20K | AP50: 4.5 -> 6.63 | Helped |
| Use all region proposals instead of 400 (ablation) | LOD · COCO C20K | CorLoc: 48.0 -> 48.5 | Helped |
| Use self-supervised instead of supervised features and proposals (ablation) | LOD · COCO C20K | CorLoc: 48.5 -> 41.1 | Hurt |
| Use self-supervised instead of supervised features and proposals (ablation) | LOD · Op50K | CorLoc: 48.1 -> 49.5 | Helped |
| Use VGG19 instead of VGG16 features (ablation) | LOD · Op50K | CorLoc: 48.1 -> 45.1 | Hurt |
| Use ResNet50 instead of VGG16 features (ablation) | LOD · COCO C20K | CorLoc: 48.5 -> 35.4 | Hurt |
| Increase PageRank damping beta from 10^-4 to 10^-1 (ablation) | PageRank · COCO C20K | CorLoc: 48.0 -> 40.0 | Hurt |
| Vary selection factor alpha from 0.05 to 0.2 (ablation) | LOD · COCO C20K | CorLoc: 48.4 -> 48.5 | No clear effect |
Closest papers by meaning
- unMORE: Unsupervised Multi-Object Segmentation via Center-Boundary Reasoning (ICML 2025 · similarity 0.87)
- CuVLER: Enhanced Unsupervised Object Discoveries through Exhaustive Self-Supervised Transformers (CVPR 2024 · similarity 0.86)
- UN-DETR: Promoting Objectness Learning via Joint Supervision for Unknown Object Detection (AAAI 2025 · similarity 0.86)
- UNION: Unsupervised 3D Object Detection using Object Appearance-based Pseudo-Classes (NeurIPS 2024 · similarity 0.86)
- Multi-Part Object Representations via Graph Structures and Co-Part Discovery (NeurIPS 2026 · similarity 0.86)
- Pose-Guided Self-Training with Two-Stage Clustering for Unsupervised Landmark Discovery (CVPR 2024 · similarity 0.85)
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.