Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Unsupervised Few-Shot Image Classification by Learning Features into Clustering Space
Shuo Li, Fang Liu, Zehua Hao, Kaibo Zhao, Licheng Jiao
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 |
|---|---|---|---|
| build tasks from LF2CS clusters instead of clustering-based baseline | unsupervised few-shot (ResNet-12) · miniImageNet 5-way 1-shot | accuracy: 48.48 -> 53.14 | Improved |
| build tasks from LF2CS clusters instead of CSS-based augmentation views | unsupervised few-shot (ResNet-12) · miniImageNet 5-way 1-shot | accuracy: 45.42 -> 53.14 | Improved |
| build tasks from LF2CS clusters instead of random selection | unsupervised few-shot (ResNet-12) · miniImageNet 5-way 1-shot | accuracy: 29.92 -> 53.14 | Improved |
| use LF2CS instead of clustering-based task building | unsupervised few-shot (Conv-4) · Omniglot 5-way 1-shot | accuracy: 83.14 -> 97.31 | Improved |
| use LF2CS instead of clustering-based task building | unsupervised few-shot (ResNet-12) · tieredImageNet 5-way 1-shot | accuracy: 44.93 -> 53.16 | Improved |
| use LF2CS instead of clustering-based task building | unsupervised few-shot (ResNet-12) · CIFARFS 5-way 1-shot | accuracy: 44.88 -> 55.04 | Improved |
| use LF2CS instead of CSS-based task building | unsupervised few-shot (Conv-4) · CIFARFS 5-way 1-shot | accuracy: 42.59 -> 51.52 | Improved |
| use ResNet-12 encoder instead of Conv-4 (ablation) | LF2CS · miniImageNet 5-way 1-shot | accuracy: 48.32 -> 53.14 | Improved |
| use ResNet-12 encoder instead of Conv-4 (ablation) | LF2CS · miniImageNet test KNN top-1 | accuracy: 49.10 -> 60.25 | Improved |
| increase LF2CS head dimension d (ablation) | LF2CS · miniImageNet KNN top-1 | accuracy | Improved |
| evaluate on novel test classes instead of training classes (ablation) | LF2CS (Conv-4) · miniImageNet 5-way 1-shot | accuracy: 48.41 -> 48.32 | No clear effect |
| evaluate on novel test classes instead of training classes (ablation) | LF2CS (ResNet-12) · miniImageNet 5-way 1-shot | accuracy: 53.59 -> 53.14 | No clear effect |
Closest papers by meaning
- SCAN: Learning to Classify Images without Labels (ECCV 2020 · similarity 0.91)
- Invariant Information Clustering for Unsupervised Image Classification and Segmentation (ICCV 2019 · similarity 0.89)
- Semi-Supervised Few-Shot Learning via Multi-Factor Clustering (CVPR 2022 · similarity 0.88)
- Meta-Learning for Semi-Supervised Few-Shot Classification (ICLR 2018 · similarity 0.88)
- Mixture-Based Feature Space Learning for Few-Shot Image Classification (ICCV 2021 · similarity 0.88)
- Discriminative Sample-Guided and Parameter-Efficient Feature Space Adaptation for Cross-Domain Few-Shot Learning (CVPR 2024 · 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.
