Research map / Continual learning / Class-incremental
Uncertainty-Aware Contrastive Distillation for Incremental Semantic Segmentation
Guanglei Yang, Enrico Fini, Dan Xu, Paolo Rota, Mingli Ding, Moin Nabi, Xavier Alameda-Pineda, Elisa Ricci
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 UCD on top of MiB | MiB · VOC2012 15-1 overlapped | mIoU: 29.7 -> 42.2 | Improved |
| add UCD on top of MiB | MiB · VOC2012 19-1 disjoint | mIoU: 67.4 -> 72.0 | Improved |
| add UCD on top of PLOP | PLOP · VOC2012 15-5 disjoint | mIoU: 59.8 -> 60.1 | Improved |
| add UCD on top of PLOP | PLOP · Cityscapes 13-6 overlapped | mIoU: 39.6 -> 39.7 | No clear effect |
| add UCD on top of MiB | MiB · ADE20K 100-50 disjoint | mIoU: 34.6 -> 36.4 | Improved |
| add segment contrastive loss to baseline (ablation) | baseline · VOC2012 15-5 disjoint | mIoU: 59.4 -> 61.4 | Improved |
| use pixel contrastive instead of segment contrastive (ablation) | baseline · VOC2012 15-5 disjoint | mIoU: 61.4 -> 62.9 | Improved |
| use contrastive distillation instead of pixel contrastive (ablation) | baseline · VOC2012 15-5 disjoint | mIoU: 62.9 -> 65.3 | Improved |
| add uncertainty estimation to contrastive distillation (ablation) | UCD · VOC2012 15-5 disjoint | mIoU: 65.3 -> 66.3 | Improved |
| add uncertainty estimation to contrastive distillation (ablation) | UCD · ADE20K 100-50 disjoint | mIoU: 35.8 -> 36.4 | Improved |
| use temperature 0.07 instead of 1 (ablation) | UCD · VOC2012 15-5 disjoint | mIoU: 60.5 -> 66.3 | Improved |
| exclude background from contrastive loss (ablation) | UCD · VOC2012 | mIoU: 66.9 -> 66.5 | No clear effect |
Closest papers by meaning
- Inherit With Distillation and Evolve With Contrast: Exploring Class Incremental Semantic Segmentation Without Exemplar Memory (TPAMI 2023 · similarity 0.93)
- Augmentation-Free Dense Contrastive Knowledge Distillation for Efficient Semantic Segmentation (NeurIPS 2023 · similarity 0.90)
- Modeling the Background for Incremental Learning in Semantic Segmentation (CVPR 2020 · similarity 0.90)
- Decomposed Knowledge Distillation for Class-Incremental Semantic Segmentation (NeurIPS 2022 · similarity 0.89)
- Class Similarity Weighted Knowledge Distillation for Continual Semantic Segmentation (CVPR 2022 · similarity 0.89)
- Alleviating Catastrophic Forgetting of Incremental Object Detection via Within-Class and Between-Class Knowledge Distillation (ICCV 2023 · similarity 0.89)
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.
