Research map / Continual learning / Class-incremental
Strike a Balance in Continual Panoptic Segmentation
Jinpeng Chen, Runmin Cong, Yuxuan Luo, Horace Ho Shing Ip, Sam Kwong
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 past-class backtrace distillation (ablation) | Mask2Former with pseudo-labels · ADE20K CPS 100-10 | PQ (all): 32.0 -> 34.7 | Improved |
| distill only past-class features instead of entire decoder features (ablation) | BalConpas · ADE20K CPS 100-10 | PQ (all): 33.8 -> 34.7 | Improved |
| add entire-feature distillation vs no distillation (ablation) | Mask2Former with pseudo-labels · ADE20K CPS 100-10 | PQ (all): 32.0 -> 33.8 | Improved |
| use class-proportional memory instead of random replay selection (ablation) | BalConpas · ADE20K CPS 100-10 | PQ (all): 35.2 -> 36.5 | Improved |
| use class-proportional memory instead of SSUL equal-per-class selection (ablation) | BalConpas · ADE20K CPS 100-10 | PQ (all): 35.8 -> 36.5 | Improved |
| add balanced anti-misguidance losses to replay (ablation) | BalConpas with CPM · ADE20K CPS 100-10 | PQ (all): 35.1 -> 36.5 | Improved |
| base class distribution on segment counts instead of pixel counts (ablation) | CPM · ADE20K CPS 100-10 | PQ (all): 35.2 -> 36.5 | Improved |
| base class distribution on segment counts instead of image counts (ablation) | CPM · ADE20K CPS 100-10 | PQ (all): 36.0 -> 36.5 | Improved |
| add any replay selection strategy (random replay with BAG) (ablation) | BalConpas without replay · ADE20K CPS 100-10 | PQ (all): 34.7 -> 35.2 | Improved |
| add replay with CPM and BAG to BalConpas | BalConpas · ADE20K CPS 100-5 | PQ (all): 30.8 -> 32.6 | Improved |
| BalConpas versus EWF on average mIoU | BalConpas · ADE20K CSS 100-5 | mIoU (avg): 43.4 -> 43.1 | Hurt |
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.
- CoMFormer — CoMFormer: Continual Learning in Semantic and Panoptic Segmentation (CVPR 2023) · better on 3 of 3 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
- ECLIPSE: Efficient Continual Learning in Panoptic Segmentation with Visual Prompt Tuning (CVPR 2024 · similarity 0.88)
- Beyond Background Shift: Rethinking Instance Replay in Continual Semantic Segmentation (CVPR 2025 · similarity 0.87)
- Towards Continual Universal Segmentation (CVPR 2025 · similarity 0.87)
- Adaptive Prototype Replay for Class Incremental Semantic Segmentation (AAAI 2025 · similarity 0.87)
- Rethinking Query-based Transformer for Continual Image Segmentation (CVPR 2025 · similarity 0.86)
- Continual Segmentation with Disentangled Objectness Learning and Class Recognition (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.
