Research map / Detection and segmentation / Weakly supervised detection and segmentation
Self-Supervised Transformers for Unsupervised Object Discovery Using Normalized Cut
Check 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 TokenCut's graph-based normalized cut instead of LOST's seed expansion | ViT-S/16 object discovery · VOC07 | CorLoc: 61.9 -> 68.8; +6.9 | Helped |
| Use TokenCut's graph-based normalized cut instead of LOST's seed expansion | ViT-S/16 object discovery · VOC12 | CorLoc: 64.0 -> 72.1; +8.1 | Helped |
| Use TokenCut's graph-based normalized cut instead of LOST's seed expansion | ViT-S/16 object discovery · COCO20K | CorLoc: 50.7 -> 58.8; +8.1 | Helped |
| Add a second-stage class-agnostic detector | TokenCut · VOC07 | CorLoc: 65.7 -> 71.4; +5.7 | Helped |
| Add Bilateral Solver post-processing | TokenCut · DUT-OMRON | IoU: 53.3 -> 61.8; +12.9 | Helped |
| Vary the similarity threshold tau across 0, 0.1, 0.2, and 0.3 (ablation) | TokenCut · VOC07 | CorLoc: 67.4; 68.6; 68.8; 67.7 | No clear effect |
| Replace DINO pretraining with MoCoV3 (ablation) | TokenCut · VOC07 | CorLoc: 68.8 -> 66.2 | Hurt |
| Use patch size 8 instead of 16 (ablation) | TokenCut with ViT-S · VOC07 | CorLoc: 68.8 -> 67.3 | Hurt |
| Replace mean thresholding with EM clustering of the eigenvector (ablation) | TokenCut · VOC07 | CorLoc: 68.8 -> 63.0 | Hurt |
| Replace mean thresholding with K-means clustering of the eigenvector (ablation) | TokenCut · VOC07 | CorLoc: 68.8 -> 67.5 | Hurt |
Closest papers by meaning
- CuVLER: Enhanced Unsupervised Object Discoveries through Exhaustive Self-Supervised Transformers (CVPR 2024 · similarity 0.88)
- Unsupervised Object Localization: Observing the Background To Discover Objects (CVPR 2023 · similarity 0.88)
- Unsupervised Semantic Segmentation with Self-supervised Object-centric Representations (ICLR 2023 · similarity 0.88)
- UNION: Unsupervised 3D Object Detection using Object Appearance-based Pseudo-Classes (NeurIPS 2024 · similarity 0.86)
- Self-Supervised Learning of Intertwined Content and Positional Features for Object Detection (ICML 2025 · similarity 0.86)
- HEAP: Unsupervised Object Discovery and Localization with Contrastive Grouping (AAAI 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.