Research map / Continual learning / Class-incremental
Learning Yourself: Class-Incremental Semantic Segmentation with Language-Inspired Bootstrapped Disentanglement
Ruitao Wu, Yifan Zhao, Jia Li
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 learnable prompts (ablation) | CLIP-based CISS baseline · Pascal VOC 15-1 (overlapped) | mIoU (All): 68.9 -> 74.8 | Improved |
| add language-guided prototypical disentangle (LPD) (ablation) | prompt-tuned CLIP · Pascal VOC 15-1 (overlapped) | mIoU (All): 74.8 -> 77.4 | Improved |
| add manifold background representation (ablation) | prompt + LPD · Pascal VOC 15-1 (overlapped) | mIoU (All): 77.4 -> 77.7 | Improved |
| add mutual background disentangle (MBD) (ablation) | prompt + LPD + manifold · Pascal VOC 15-1 (overlapped) | mIoU (All): 77.7 -> 78.1 | Improved |
| add mutual background disentangle (MBD) (ablation) | prompt + LPD + manifold · Pascal VOC 15-1 (overlapped) | mIoU (16-20): 66.1 -> 66.6 | Improved |
| replace ViT backbone with CLIP in INC (ablation) | INC · Pascal VOC 15-1 (overlapped) | mIoU (All): 75.6 -> 76.1 | Improved |
| replace ViT backbone with CLIP in MBS (ablation) | MBS · Pascal VOC 15-1 (overlapped) | mIoU (All): 77.6 -> 76.8 | Hurt |
| use LBD instead of fixed variant (ablation) | LBD · Pascal VOC 15-1 (overlapped) | mIoU (16-20): 49.0 -> 66.6 | Improved |
| train prompts only (PEFT) instead of full encoder fine-tuning (ablation) | LBD · Pascal VOC 15-1 (overlapped) | mIoU (All): 78.1 -> 64.8 | Hurt |
| train prompts and adapter (PEFT) instead of full encoder (ablation) | LBD · Pascal VOC 15-1 (overlapped) | mIoU (All): 78.1 -> 72.1 | Hurt |
| train prompts and image encoder without adapter (ablation) | LBD · Pascal VOC 15-1 (overlapped) | mIoU (All): 78.1 -> 77.4 | Hurt |
| use LBD instead of MBS on ADE20k 100-50 | LBD · ADE20k 100-50 (overlapped) | mIoU (All): 45.3 -> 47.1 | Improved |
Closest papers by meaning
- Incrementer: Transformer for Class-Incremental Semantic Segmentation With Knowledge Distillation Focusing on Old Class (CVPR 2023 · similarity 0.89)
- Modeling the Background for Incremental Learning in Semantic Segmentation (CVPR 2020 · similarity 0.89)
- Cs2K: Class-specific and Class-shared Knowledge Guidance for Incremental Semantic Segmentation (ECCV 2024 · similarity 0.88)
- Deconfound Semantic Shift and Incompleteness in Incremental Few-shot Semantic Segmentation (AAAI 2025 · similarity 0.87)
- CoMBO: Conflict Mitigation via Branched Optimization for Class Incremental Segmentation (CVPR 2025 · similarity 0.87)
- Reducing Class-wise Confusion for Incremental Learning with Disentangled Manifolds (CVPR 2025 · 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.
