Research map / Detection and segmentation / Open-set and few-shot detection
CrOC: Cross-View Online Clustering for Dense Visual 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 |
|---|---|---|---|
| Use CrOC rather than DenseCL for unsupervised segmentation | frozen dense features · PVOC12 | mIoU: 18.0 vs 16.1 | No clear effect |
| Use CrOC rather than DenseCL for unsupervised segmentation | frozen dense features · COCO-Stuff | mIoU: 16.9 -> 21.9 | Helped |
| Increase positional-cue weight from 0 to 4 (ablation) | CrOC · PVOC12 unsupervised segmentation | mIoU: 3.4 -> 16.4 | Helped |
| Increase positional-cue weight from 0 to 4 (ablation) | CrOC · PVOC12 linear segmentation | mIoU: 52.2 -> 55.0 | Mixed |
| Use positional-cue weight 4 rather than 0 (ablation) | CrOC · COCO-Things unsupervised segmentation | mIoU: 6.8 -> 17.0 | Helped |
| Use positional-cue weight 4 rather than 0 (ablation) | CrOC · COCO-Stuff unsupervised segmentation | mIoU: 6.3 -> 20.9 | Helped |
| Increase initial centroid count from 4 to 16 (ablation) | CrOC · PVOC12 linear segmentation | mIoU: 48.0 -> 56.9 | Helped |
| Set initial centroid count to 8 rather than 12 (ablation) | CrOC · COCO-Stuff unsupervised segmentation | mIoU: 17.9 -> 23.4 | Helped |
| Use values tokens rather than last spatial tokens for clustering (ablation) | CrOC · PVOC12 unsupervised segmentation | mIoU: 11.2 -> 16.5 | Helped |
| Use keys tokens rather than queries for clustering (ablation) | CrOC · PVOC12 unsupervised segmentation | mIoU: 8.3 -> 15.7 | Helped |
| Pre-train CrOC on ImageNet rather than COCO+ | CrOC · PVOC12 linear segmentation | mIoU: 60.6 -> 70.6 | Helped |
| Pre-train CrOC on ImageNet rather than COCO+ | CrOC · PVOC12 unsupervised segmentation | mIoU: 20.6 -> 3.8 | Hurt |
Closest papers by meaning
- ClusterFomer: Clustering As A Universal Visual Learner (NeurIPS 2023 · similarity 0.86)
- MuM: Multi-View Masked Image Modeling for 3D Vision (CVPR 2026 · similarity 0.86)
- unMORE: Unsupervised Multi-Object Segmentation via Center-Boundary Reasoning (ICML 2025 · similarity 0.86)
- CrIBo: Self-Supervised Learning via Cross-Image Object-Level Bootstrapping (ICLR 2024 · similarity 0.86)
- Hierarchical Compact Clustering Attention (COCA) for Unsupervised Object-Centric Learning (CVPR 2025 · similarity 0.85)
- Segment Anything without Supervision (NeurIPS 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.