Research map / Detection and segmentation / Open-set and few-shot detection
Learning Accurate Segmentation Purely from Self-Supervision
Zuyao You, Zuxuan Wu, Yu-Gang Jiang
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 Iterative Patch Optimization to NCut (ablation) | Selfment initial mask · ECSSD | Fmax / IoU / Acc: 74.7 -> 79.5 Fmax, 63.9 -> 73.2 IoU, 86.2 -> 87.8 Acc | Improved |
| add self-supervised head training with BCE on IPO pseudo-labels (ablation) | Selfment · ECSSD | Fmax: 79.5 -> 88.3 | Improved |
| add contrastive loss to BCE (ablation) | Selfment head training · ECSSD | Fmax / IoU: 88.3 -> 88.9 Fmax, 80.4 -> 81.4 IoU | Improved |
| add Dice loss to BCE (ablation) | Selfment head training · ECSSD | Fmax / IoU: 88.3 -> 88.9 Fmax, 80.4 -> 81.3 IoU | Improved |
| add both Dice and contrastive loss to BCE (ablation) | Selfment head training · ECSSD | Fmax: 88.3 -> 89.1 | Improved |
| NCut instead of <CLS> token as initial bipartition (ablation) | Selfment initialization · ECSSD | Fmax: 29.2 -> 74.7 | Improved |
| NCut instead of K-means as initial bipartition (ablation) | Selfment initialization · ECSSD | Fmax: 59.2 -> 74.7 | Improved |
| increase inference resolution from 768 to 1280 (ablation) | Selfment · ECSSD | Fmax: 95.3 -> 95.9 | Improved |
| increase inference resolution from 768 to 1280 (ablation) | TokenCut · ECSSD | Fmax: 87.8 -> 86.7 | Hurt |
| increase inference resolution from 768 to 1280 (ablation) | SelfMask · ECSSD | Fmax: 91.9 -> 89.9 | Hurt |
| increase inference resolution from 768 to 1280 (ablation) | FOUND · ECSSD | Fmax: 91.4 -> 89.2 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- TokenCut — TokenCut: Segmenting Objects in Images and Videos With Self-Supervised Transformer and Normalized Cut (TPAMI 2023) · better on 4 of 4 shared results
- UCOD-DPL — UCOD-DPL: Unsupervised Camouflaged Object Detection via Dynamic Pseudo-label Learning (CVPR 2025) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Self-Supervised Learning of Object Parts for Semantic Segmentation (CVPR 2022 · similarity 0.90)
- Rethinking Self-Supervised Semantic Segmentation: Achieving End-to-End Segmentation (TPAMI 2024 · similarity 0.90)
- Self-Supervised Human Detection and Segmentation via Background Inpainting (TPAMI 2021 · similarity 0.90)
- Segment Anything without Supervision (NeurIPS 2024 · similarity 0.88)
- Object Discovery and Representation Networks (ECCV 2022 · similarity 0.88)
- The Devil is in the Object Boundary: Towards Annotation-free Instance Segmentation using Foundation Models (ICLR 2024 · 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.
