Research map / Continual learning / Class-incremental
Class-Incremental Learning: Survey and Performance Evaluation on Image Classification
Marc Masana, Xialei Liu, Bartłomiej Twardowski, Mikel Menta, Andrew D. Bagdanov, Joost van de Weijer
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 |
|---|---|---|---|
| expand exemplar memory from 10 to 20 per class (ablation) | BiC · CIFAR-100 (10/10) | average accuracy: 6.2 point gain | Improved |
| expand exemplar memory from 20 to 40 per class (ablation) | BiC · CIFAR-100 (10/10) | average accuracy: 4.8 point gain | Improved |
| add exemplar memory to finetuning (FT-E vs FT) (ablation) | Finetuning · CIFAR-100 | average accuracy | Improved |
| freeze weights of previous-class outputs (FT+ vs FT) (ablation) | Finetuning · CIFAR-100 | average accuracy | Improved |
| add exemplars to freezing baseline (FZ+ to FZ-E) (ablation) | Freezing · CIFAR-100 | average accuracy | Hurt |
| add exemplars to LwF (LwF-E) compared to finetuning with exemplars (ablation) | LwF · CIFAR-100 (10/10) | average accuracy | Hurt |
| use NME classifier (iCaRL) instead of finetuning with exemplars | FT-E · CIFAR-100 (5/20, ResNet-32 from scratch, 2000 exemplars) | average accuracy: 40.9 -> 43.5 | Improved |
| add balanced finetuning on exemplars (EEIL) to FT-E | FT-E · CIFAR-100 (5/20, ResNet-32 from scratch, 2000 exemplars) | average accuracy: 40.9 -> 47.6 | Improved |
Closest papers by meaning
- Class-Incremental Learning: A Survey (TPAMI 2024 · similarity 0.95)
- IL2M: Class Incremental Learning With Dual Memory (ICCV 2019 · similarity 0.92)
- Class-Incremental Learning With Strong Pre-Trained Models (CVPR 2022 · similarity 0.92)
- Class-Incremental Learning by Knowledge Distillation With Adaptive Feature Consolidation (CVPR 2022 · similarity 0.91)
- A Model or 603 Exemplars: Towards Memory-Efficient Class-Incremental Learning (ICLR 2023 · similarity 0.91)
- Revisiting Generative Replay for Class Incremental Object Detection (CVPR 2025 · 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.
