Research map / Continual learning / Class-incremental
Category Adaptation Meets Projected Distillation in Generalized Continual Category Discovery
Grzegorz Rypeść, Daniel Marczak, Sebastian Cygert, Tomasz Trzcinski, Bartlomiej Twardowski
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 projected distillation plus centroid adaptation (CAMP) vs best prior exemplar-free method | CAMP · CUB200 GCCD | All accuracy: 50.3 -> 58.9 | Improved |
| add centroid adaptation (ablation) | CAMP · CUB200 known accuracy | Known accuracy: 36.5 -> 62.6 | Improved |
| add projected distillation loss (ablation) | CAMP · CUB200 known accuracy | Known accuracy: 30.9 -> 62.6 | Improved |
| add supervised loss L_SL (ablation) | CAMP · CUB200 novel accuracy | Novel accuracy: 12.1 -> 44.2 | Improved |
| add self-supervised loss L_SSL (ablation) | CAMP · CUB200 | Known vs novel accuracy: known 64.6 -> 62.6, novel 41.4 -> 44.2 | Mixed |
| replace MLP projector distiller with plain feature distillation (ablation) | CAMP · StanfordCars GCCD | All accuracy: 38.1 -> 28.5 | Hurt |
| add centroid adaptation on top of feature distillation | GCD with FD · CIFAR100 GCCD | All accuracy | No clear effect |
| add MLP projector distillation alone without adaptation (ablation) | GCD · CIFAR100 GCCD | All accuracy: +4.7% points | Improved |
| add MLP distiller combined with linear adapter (ablation) | GCD · CIFAR100 GCCD | All accuracy: +20.2% points | Improved |
| add exemplars (20 per class) to GCD (ablation) | GCD · CUB200 GCCD | All accuracy: +16.0% (5 exemplars), +31.4% (20 exemplars) points | Improved |
| CAMP vs GCD+FD on DomainNet novel classes | CAMP · DomainNet GCCD | Novel accuracy: 39.7 vs 29.7 | Hurt |
| apply CAMP in exemplar-free class-incremental learning vs FeTrIL | CAMP · CIFAR-100 CIL N=10 | average incremental accuracy: 46.3 -> 56.7 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MetaGCD — MetaGCD: Learning to Continually Learn in Generalized Category Discovery (ICCV 2023) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Continual Generalized Category Discovery: Learning and Forgetting from a Bayesian Perspective (ICML 2025 · similarity 0.90)
- Happy: A Debiased Learning Framework for Continual Generalized Category Discovery (NeurIPS 2024 · similarity 0.89)
- Adaptive Discovering and Merging for Incremental Novel Class Discovery (AAAI 2024 · similarity 0.88)
- Solving the Catastrophic Forgetting Problem in Generalized Category Discovery (CVPR 2024 · similarity 0.86)
- When Domain Generalization meets Generalized Category Discovery: An Adaptive Task-Arithmetic Driven Approach (CVPR 2025 · similarity 0.86)
- Consistent Supervised-Unsupervised Alignment for Generalized Category Discovery (NeurIPS 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.
