Research map / Continual learning / Class-incremental
Few-Shot Class-Incremental Learning from an Open-Set Perspective
Can Peng, Kun Zhao, Tianren Wang, Meng Li, Brian C. Lovell
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 |
|---|---|---|---|
| replace cross-entropy with angular penalty (cosFace) loss (ablation) | ALICE (balanced prototypes) · CIFAR100 5-shot, last session | average accuracy: 46.2 -> 49.5 | Improved |
| replace cross-entropy with angular penalty (cosFace) loss (ablation) | ALICE (balanced prototypes) · CIFAR100 5-shot, last session | harmonic accuracy: 31.4 -> 25.2 | Hurt |
| add projection layers before the final fully connected layer (ablation) | ALICE · CIFAR100 5-shot, last session | harmonic accuracy: 25.2 -> 43.8 | Improved |
| add projection layers (ablation) | ALICE · CIFAR100 5-shot, last session | average accuracy: 49.5 -> 49.5 | No clear effect |
| add two augmentations of each input image (data augmentation) (ablation) | ALICE · CIFAR100 5-shot, last session | average accuracy: 49.5 -> 51.3 | Improved |
| add two augmentations of each input image (data augmentation) (ablation) | ALICE · CIFAR100 5-shot, last session | harmonic accuracy: 43.8 -> 47.4 | Improved |
| add class augmentation by mixing class pairs (ablation) | ALICE · CIFAR100 5-shot, last session | average accuracy: 51.3 -> 54.1 | Improved |
| add class augmentation by mixing class pairs (ablation) | ALICE · CIFAR100 5-shot, last session | harmonic accuracy: 47.4 -> 50.6 | Improved |
| use balanced data (same number of base samples) for prototypes (ablation) | cross-entropy model · CIFAR100 5-shot, last session | harmonic accuracy: 27.4 -> 31.4 | Improved |
| use balanced data for prototypes (ablation) | angular penalty model · CIFAR100 5-shot, last session | harmonic accuracy: 22.3 -> 25.2 | Improved |
| add class augmentation on fine-grained data (ablation) | ALICE · CUB200 | accuracy | Hurt |
| replace CEC with ALICE | CEC · CIFAR100 1-shot, last session | harmonic accuracy: 13.0 -> 26.5 | 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.
- iCaRL — iCaRL: Incremental Classifier and Representation Learning (CVPR 2017) · 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
- Few-Shot Class-Incremental Learning for Classification and Object Detection: A Survey (TPAMI 2025 · similarity 0.92)
- Rethinking Few-shot Class-incremental Learning: Learning from Yourself (ECCV 2024 · similarity 0.92)
- CLOSER: Towards Better Representation Learning for Few-Shot Class-Incremental Learning (ECCV 2024 · similarity 0.91)
- Tripartite Weight-Space Ensemble for Few-Shot Class-Incremental Learning (CVPR 2025 · similarity 0.91)
- Adaptive Decision Boundary for Few-Shot Class-Incremental Learning (AAAI 2025 · similarity 0.90)
- OrCo: Towards Better Generalization via Orthogonality and Contrast for Few-Shot Class-Incremental Learning (CVPR 2024 · similarity 0.88)
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.
