Research map / Detection and segmentation / Open-set and few-shot detection
Toward Unsupervised, Multi-Object Discovery in Large-Scale Image Collections
Huy V. Vo, Patrick Pérez, Jean Ponce
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 one-region-per-group regularization (rOSD) to OSD (ablation) | OSD with our proposals · VOC_all colocalization (VGG16) | CorLoc: 44.7 -> 45.8 | Improved |
| add one-region-per-group regularization (rOSD) to OSD (ablation) | OSD with our proposals · VOC_all single-object discovery (VGG16) | CorLoc: 48.7 -> 49.2 | Improved |
| add one-region-per-group regularization (rOSD) to OSD (ablation) | OSD with our proposals · OD and VOC_6x2 colocalization (VGG16) | CorLoc: 89.0 -> 89.0; 73.6 -> 73.3 | No clear effect |
| add one-region-per-group regularization (rOSD) to OSD (ablation) | OSD with our proposals · VOC_all multi-object discovery (VGG19) | detection rate: 34.8 -> 37.6 | Improved |
| add one-region-per-group regularization (rOSD) to OSD (ablation) | OSD with our proposals · VOC12 multi-object discovery (VGG19) | detection rate: 36.8 -> 40.4 | Improved |
| use our unsupervised CNN proposals instead of randomized Prim (ablation) | OSD · VOC_all colocalization (VGG16) | CorLoc: 42.8 -> 44.7 | Improved |
| use our unsupervised CNN proposals instead of selective search (ablation) | OSD · VOC_all colocalization (VGG16) | CorLoc: 30.9 -> 44.7 | Improved |
| add two-stage large-scale strategy (ablation) | rOSD · VOC_all single-object discovery | CorLoc: 42.8 -> 49.4 | Improved |
| add two-stage large-scale strategy (ablation) | rOSD · COCO_20k single-object discovery | CorLoc: 44.5 -> 48.5 | Improved |
| add two-stage large-scale strategy (ablation) | rOSD · COCO_20k multi-object discovery | detection rate: 11.6 -> 12.0 | Improved |
| add ensemble method of Vo et al. (ablation) | OSD with our proposals · VOC_6x2 colocalization (VGG16) | CorLoc: 73.6 -> 75.3 | Mixed |
| use standout score instead of confidence score (ablation) | OSD with our proposals · OD vs VOC_6x2 colocalization | CorLoc: 83.7 -> 89.0; 73.6 -> 64.1 | Mixed |
Closest papers by meaning
- Large-Scale Unsupervised Object Discovery (NeurIPS 2021 · similarity 0.90)
- Spatially Invariant Unsupervised Object Detection with Convolutional Neural Networks (AAAI 2019 · similarity 0.89)
- Unsupervised Image Matching and Object Discovery as Optimization (CVPR 2019 · similarity 0.89)
- HEAP: Unsupervised Object Discovery and Localization with Contrastive Grouping (AAAI 2024 · similarity 0.87)
- 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)
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.
