Research map / Continual learning / Class-incremental
Incrementer: Transformer for Class-Incremental Semantic Segmentation With Knowledge Distillation Focusing on Old Class
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 |
|---|---|---|---|
| Use Incrementer instead of RBC | Transformer-based class-incremental segmentation · Pascal VOC 15-1 overlapped | mIoU (all): 62.14 -> 75.55 | Helped |
| Use Incrementer instead of the strongest listed prior method | Transformer-based class-incremental segmentation · ADE20k 100-5 overlapped | mIoU (all): 28.75 -> 41.72 | Helped |
| Focus visual-embedding distillation on old-class regions rather than distilling all visual embeddings (ablation) | Incrementer with VKD baseline · Pascal VOC 15-1 overlapped | mIoU (all): 62.96 -> 65.82 | Helped |
| Add old-class-focused class-embedding distillation to visual FOD (ablation) | Incrementer with Vis FOD · Pascal VOC 15-1 overlapped | mIoU (all): 65.82 -> 68.23 | Helped |
| Add class deconfusion strategy to FOD (ablation) | Incrementer · Pascal VOC 15-1 overlapped | mIoU (all): 68.23 -> 75.55 | Helped |
| Reduce the segmentation loss weight for new classes (ablation) | Incrementer without CDS · Pascal VOC 15-1 overlapped | mFP (new classes): 38.76 -> 22.45 | Helped |
| Reduce the segmentation loss weight for new classes (ablation) | Incrementer without CDS · Pascal VOC 15-1 overlapped | mIoU (all): 68.23 -> 71.27 | Helped |
| Add the old-new binary-mask loss (ablation) | Incrementer without CDS · Pascal VOC 15-1 overlapped | mIoU (all): 68.23 -> 73.67 | Helped |
| Combine new-class loss reweighting with old-new binary-mask loss (ablation) | Incrementer · Pascal VOC 15-1 overlapped | mFP (new classes): 38.76 -> 26.80 | Helped |
| Combine new-class loss reweighting with old-new binary-mask loss (ablation) | Incrementer · Pascal VOC 15-1 overlapped | mIoU (all): 68.23 -> 75.55 | Helped |
Closest papers by meaning
- Decomposed Knowledge Distillation for Class-Incremental Semantic Segmentation (NeurIPS 2022 · similarity 0.94)
- Modeling the Background for Incremental Learning in Semantic Segmentation (CVPR 2020 · similarity 0.93)
- SSUL: Semantic Segmentation with Unknown Label for Exemplar-based Class-Incremental Learning (NeurIPS 2021 · similarity 0.92)
- CoMBO: Conflict Mitigation via Branched Optimization for Class Incremental Segmentation (CVPR 2025 · similarity 0.89)
- Distilling Knowledge from Heterogeneous Architectures for Semantic Segmentation (AAAI 2025 · similarity 0.88)
- Deconfound Semantic Shift and Incompleteness in Incremental Few-shot Semantic Segmentation (AAAI 2025 · 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.