Research map / Label-efficient and robust learning / Recognition and long-tail
Automatically Discovering Novel Visual Categories With Adaptive Prototype Learning
Lu Zhang, Lu Qi, Xu Yang, Hong Qiao, Ming–Hsuan Yang, Zhiyong Liu
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 instance discrimination (DINO) (ablation) | Ours · CIFAR100 | clustering accuracy: 64.7% -> 78.9% | Improved |
| add instance discrimination (DINO) (ablation) | Ours · CIFAR10 | clustering accuracy: 91.0% -> 96.0% | Improved |
| add instance discrimination (DINO) (ablation) | Ours · OmniGlot | clustering accuracy: 89.5% -> 93.4% | Improved |
| add category discrimination (online prototype learning) (ablation) | Ours · CIFAR10 | clustering accuracy: 92.9% -> 96.0% | Improved |
| add category discrimination (online prototype learning) (ablation) | Ours · CIFAR100 | clustering accuracy: 72.8% -> 78.9% | Improved |
| add category discrimination (online prototype learning) (ablation) | Ours · OmniGlot | clustering accuracy: 91.2% -> 93.4% | Improved |
| add prototypical self-training (stage II) (ablation) | Ours · CIFAR10 | clustering accuracy: 93.6% -> 96.0% | Improved |
| add prototypical self-training (stage II) (ablation) | Ours · CIFAR100 | clustering accuracy: 75.6% -> 78.9% | Improved |
| add OPL to k-means pseudo labels (ablation) | Ours · CIFAR10 | clustering accuracy: 91.1% -> 93.6% | Improved |
| use offline k-means instead of online prototype labels (ablation) | Ours · CIFAR10 | clustering accuracy: 90.5% -> 93.6% | Improved |
| replace SimCLR/MoCo with DINO as self-supervised method (ablation) | Ours · CIFAR100 | clustering accuracy: 61.8% -> 78.9% | Improved |
| add the proposed method on top of DINO k-means (ablation) | DINO · CIFAR100 | clustering accuracy: 69.6% -> 78.9% | 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.
- AutoNovel — AutoNovel: Automatically Discovering and Learning Novel Visual Categories (TPAMI 2021) · better on 4 of 4 shared results
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
- Semantic-Guided Novel Category Discovery (AAAI 2024 · similarity 0.91)
- Automatically Discovering and Learning New Visual Categories with Ranking Statistics (ICLR 2020 · similarity 0.91)
- AutoNovel: Automatically Discovering and Learning Novel Visual Categories (TPAMI 2021 · similarity 0.91)
- ProtoGCD: Unified and Unbiased Prototype Learning for Generalized Category Discovery (TPAMI 2025 · similarity 0.90)
- Labeled Data Selection for Category Discovery (ECCV 2024 · similarity 0.90)
- Self-Organizing Visual Prototypes for Non-Parametric Representation Learning (ICML 2025 · similarity 0.87)
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.
