Research map / Continual learning / Class-incremental
Replay Master: Automatic Sample Selection and Effective Memory Utilization for Continual Semantic Segmentation
Lanyun Zhu, Tianrun Chen, Jianxiong Yin, Simon Chong Wee See, De Wen Soh, Jun Liu
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 RL-based automatic sample selection instead of random selection (ablation) | RPMaster · VOC 2012 15-1 (overlapped) | mIoU (all classes): 67.78 -> 73.85 | Improved |
| add effective memory utilization (ablation) | RPMaster · VOC 2012 15-1 (overlapped) | mIoU (all classes): 71.94 -> 73.85 | Improved |
| add sample enhancement after selection (ablation) | RPMaster · VOC 2012 15-1 (overlapped) | mIoU (all classes): 71.31 -> 73.85 | Improved |
| add sample diversity to the state (ablation) | RPMaster · VOC 2012 15-1 (overlapped) | mIoU (all classes): 69.05 -> 73.85 | Improved |
| add accuracy I to the state (ablation) | RPMaster · VOC 2012 15-1 (overlapped) | mIoU (all classes): 71.55 -> 73.85 | Improved |
| add forgetfulness g to the state (ablation) | RPMaster · VOC 2012 15-1 (overlapped) | mIoU (all classes): 71.79 -> 73.85 | Improved |
| use multi-structure graph similarity instead of prototype-level (ablation) | RPMaster · VOC 2012 15-1 (overlapped) | mIoU (all classes): 71.19 -> 73.85 | Improved |
| use pixel-level similarity instead of graph similarity (ablation) | RPMaster · VOC 2012 15-1 (overlapped) | mIoU (all classes): 73.85 -> 74.14 | Mixed |
| add per-class experts (ablation) | RPMaster · VOC 2012 15-1 (overlapped) | mIoU (all classes): 72.05 -> 73.85 | Improved |
| add class-balanced finetuning phase (ablation) | RPMaster · VOC 2012 15-1 (overlapped) | mIoU (all classes): 72.97 -> 73.85 | Improved |
| add shared information utilization across experts (ablation) | RPMaster · VOC 2012 15-1 (overlapped) | mIoU (all classes): 72.87 -> 73.85 | Improved |
| add random interruption tau (ablation) | RPMaster · VOC 2012 15-1 (overlapped) | mIoU (all classes): 73.12 -> 73.85 | Improved |
Closest papers by meaning
- Beyond Background Shift: Rethinking Instance Replay in Continual Semantic Segmentation (CVPR 2025 · similarity 0.91)
- TIKP: Text-to-Image Knowledge Preservation for Continual Semantic Segmentation (AAAI 2024 · similarity 0.88)
- Adaptive Prototype Replay for Class Incremental Semantic Segmentation (AAAI 2025 · similarity 0.88)
- Enhancing Visual Continual Learning with Language-Guided Supervision (CVPR 2024 · similarity 0.87)
- Summarizing Stream Data for Memory-Constrained Online Continual Learning (AAAI 2024 · similarity 0.87)
- Towards Continual Universal Segmentation (CVPR 2025 · similarity 0.87)
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.
