Research map / Continual learning / Class-incremental
Mining Unseen Classes via Regional Objectness: A Simple Baseline for Incremental Segmentation
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 |
|---|---|---|---|
| add Micro mechanism (mined unseen-class clusters) to dense prediction baseline (ablation) | DP baseline · VOC 15-1 | mIoU (all): 56.0 -> 65.9 | Helped |
| add proposal classification branch to dense prediction baseline (ablation) | DP baseline · VOC 15-1 | mIoU (all): 56.0 -> 66.9 | Helped |
| add Micro mechanism on top of DP + Proposal (ablation) | MicroSeg · VOC 15-1 | mIoU (all): 66.9 -> 69.8 | Helped |
| add dense prediction branch at step 1 (ablation) | Proposal-only model · VOC 15-1 | mIoU (all): 65.8 -> 69.8 | Helped |
| vary number of future sub-classes K (1 to 9) (ablation) | MicroSeg · VOC 15-1 | mIoU (all): 67.2 (K=1) -> 69.8 (K=5) -> 68.9 (K=9) | Mixed |
| vary pseudo-label threshold tau (0.1 to 0.9) (ablation) | MicroSeg · VOC 15-1 | mIoU (all): 68.3 (0.1) to 69.8 (0.7) | No clear effect |
| add memory sampling (MicroSeg-M) to MicroSeg | MicroSeg · VOC 10-1 | mIoU (all): 61.2 -> 67.7 | Helped |
| memory sampling on ADE 50-50 | MicroSeg · ADE 50-50 | mIoU (all): 32.9 -> 31.4 | Hurt |
| MicroSeg vs SSUL without memory on long-term VOC 10-1 | SSUL · VOC 10-1 | mIoU (all): 59.3 -> 61.2 | Helped |
Closest papers by meaning
- Modeling the Background for Incremental Learning in Semantic Segmentation (CVPR 2020 · similarity 0.91)
- Deconfound Semantic Shift and Incompleteness in Incremental Few-shot Semantic Segmentation (AAAI 2025 · similarity 0.88)
- Beyond Background Shift: Rethinking Instance Replay in Continual Semantic Segmentation (CVPR 2025 · similarity 0.88)
- Continual Segmentation with Disentangled Objectness Learning and Class Recognition (CVPR 2024 · similarity 0.88)
- CoMBO: Conflict Mitigation via Branched Optimization for Class Incremental Segmentation (CVPR 2025 · similarity 0.87)
- Learning Hierarchical Image Segmentation For Recognition and By Recognition (ICLR 2024 · 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.