Research map / Continual learning / Class-incremental
CoinSeg: Contrast Inter- and Intra- Class Representations for Incremental Segmentation
Zekang Zhang, Guangyu Gao, Jianbo Jiao, Chi Harold Liu, Yunchao Wei
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 |
|---|---|---|---|
| replace ResNet101 with Swin-B backbone (freeze strategy) (ablation) | CoinSeg · Pascal VOC 2012 15-1 | mIoU (all): 63.3 -> 70.5 | Improved |
| replace freeze strategy with flexible initial learning rate only (no regularization) (ablation) | CoinSeg Swin-B · Pascal VOC 2012 15-1 | mIoU (all): 70.5 -> 66.7 | Hurt |
| add feature distillation loss to flexible learning rate (ablation) | CoinSeg · Pascal VOC 2012 15-1 | mIoU (all): 66.7 -> 70.3 | Improved |
| add logit distillation (ablation) | CoinSeg · Pascal VOC 2012 15-1 | mIoU (all): 70.3 -> 71.6 | Improved |
| add inter-class contrastive loss (ablation) | CoinSeg · Pascal VOC 2012 15-1 | mIoU (all): 71.6 -> 72.4 | Improved |
| add intra-class contrastive loss with mask proposals (ablation) | CoinSeg · Pascal VOC 2012 15-1 | mIoU (all): 72.4 -> 75.5 | Improved |
| increase number of mask proposals N from 50 to 100 to 200 (ablation) | intra-class contrastive loss · Pascal VOC 2012 15-1 | mIoU (all): 74.2 -> 75.5 -> 76.1 | Improved |
| apply flexible tuning instead of freeze to MicroSeg (ablation) | MicroSeg · Pascal VOC 2012 15-1 | mIoU (all): 71.0 -> 72.3 (ResNet101); 69.8 -> 70.4 (Swin-B) | Improved |
| use CoinSeg versus MicroSeg in few-base-class long-term scenario | class-incremental segmentation · Pascal VOC 2012 2-2 | mIoU (all): 46.5 -> 64.3 | Improved |
| add memory sampling (CoinSeg-M) (ablation) | CoinSeg · Pascal VOC 2012 15-1 | mIoU (all): 75.5 -> 79.6 | Improved |
| CoinSeg versus memory-based MicroSeg-M on novel class in 19-1 | CoinSeg-M · Pascal VOC 2012 19-1 | mIoU (class 20): 62.4 -> 52.6 | Hurt |
Later papers that compared with this method
2 later accepted papers report CoinSeg as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Learnability-Driven Knowledge Assimilation for Class-Incremental Semantic Segmentation (ICML 2026 · better on 5 of 5 shared results)
- Unlocking Positive Transfer in Incrementally Learning Surgical Instruments: A Self-reflection Hierarchical Prompt Framework (CVPR 2026 · better on 1 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SSUL — SSUL: Semantic Segmentation with Unknown Label for Exemplar-based Class-Incremental Learning (NeurIPS 2021) · 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
- Continual Gaussian Mixture Distribution Modeling for Class Incremental Semantic Segmentation (NeurIPS 2025 · similarity 0.88)
- Cs2K: Class-specific and Class-shared Knowledge Guidance for Incremental Semantic Segmentation (ECCV 2024 · similarity 0.88)
- Mining Unseen Classes via Regional Objectness: A Simple Baseline for Incremental Segmentation (NeurIPS 2022 · similarity 0.88)
- Deconfound Semantic Shift and Incompleteness in Incremental Few-shot Semantic Segmentation (AAAI 2025 · similarity 0.86)
- CoMBO: Conflict Mitigation via Branched Optimization for Class Incremental Segmentation (CVPR 2025 · similarity 0.86)
- Adaptive Prototype Replay for Class Incremental Semantic Segmentation (AAAI 2025 · 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.
