Research map / Continual learning
Class-incremental: research map
450 papers on Class-incremental in Continual learning, from ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR (2016–2026), grouped into 2 clusters and 6 approaches. The busiest year so far is 2026.
What it is
A classifier is taught new categories in stages and must recognise all classes seen so far, without being told which stage a test image comes from. The hard part is that the newest classes crowd out the old ones; methods rebalance the classifier, keep a few exemplars, or distill the old model.
Terms you will see on this page:
- Class-incremental learning (CIL)
- adding new classes over time while still recognising every earlier class.
- Exemplars
- a handful of stored images per old class used to remind the model.
- Knowledge distillation
- training the new model to match the old model's outputs so old knowledge carries over.
Start reading with the most cited papers here:
- iCaRL: Incremental Classifier and Representation Learning (CVPR 2017 · 3,848 citations)
- Large Scale Incremental Learning (CVPR 2019 · 1,269 citations)
Within Continual learning, its share shrank from 26.2% in 2023–24 to 23.4% in 2025–26 (92 → 135 papers at ICML, NeurIPS, ICLR, CVPR and AAAI, the venues with data for all four years).
Explore Class-incremental in the interactive map
Working on something in this topic? Describe your idea in scime atlas to see which approach it falls under, the closest papers by meaning and how crowded the spot has become.
What changed recently
Approaches whose share of the topic moved most between 2023–24 and 2025–26 (ICML, NeurIPS, ICLR, CVPR and AAAI):
- Incremental object detection with replay rising: 9% → 13% of the topic (8 → 17 papers)
- Few-shot class-incremental prototypes falling: 14% → 13% of the topic (13 → 17 papers)
- Exemplar-efficient class-incremental learning falling: 43% → 42% of the topic (40 → 57 papers)
Papers from 2025–2026 that are already cited most:
- Adaptive Prototype Replay for Class Incremental Semantic Segmentation (AAAI 2025 · 10 citations)
Approaches and key papers
Incremental classification and few-shot classes · 387 papers
Keywords: shot, incremental cil, feature
Approaches in this cluster:
- Exemplar-efficient class-incremental learning (154 papers)
Retains old classes with compact exemplars, embedding distillation and pretrained models. - Class-balanced knowledge retention (114 papers)
Preserves per-class knowledge and decision boundaries when new classes arrive. - Few-shot class-incremental prototypes (70 papers)
Adds new classes from few samples using prototype calibration, meta-learning and orthogonal features. - Incremental object detection with replay (49 papers)
Reduces forgetting in detectors using distillation and generative replay of old classes.
Most cited and most cited since 2024:
- iCaRL: Incremental Classifier and Representation Learning (CVPR 2017 · 3,848 citations)
- Large Scale Incremental Learning (CVPR 2019 · 1,269 citations)
- Expandable Subspace Ensemble for Pre-Trained Model-Based Class-Incremental Learning (CVPR 2024 · 117 citations)
- Gradient Reweighting: Towards Imbalanced Class-Incremental Learning (CVPR 2024 · 57 citations)
Incremental semantic segmentation · 63 papers
Keywords: segmentation, semantic, background
Approaches in this cluster:
- Unknown-class mining for incremental segmentation (36 papers)
Models background shift by mining unseen classes and using exemplar or prototype replay. - Distillation-based continual segmentation (27 papers)
Preserves old classes in segmentation using similarity-weighted distillation, instance replay and memory selection.
Most cited and most cited since 2024:
- Modeling the Background for Incremental Learning in Semantic Segmentation (CVPR 2020 · 344 citations)
- PLOP: Learning Without Forgetting for Continual Semantic Segmentation (CVPR 2021 · 265 citations)
- Continual Segmentation with Disentangled Objectness Learning and Class Recognition (CVPR 2024 · 15 citations)
- Incremental Nuclei Segmentation from Histopathological Images via Future-class Awareness and Compatibility-inspired Distillation (CVPR 2024 · 11 citations)
