Mitigating Embedding and Class Assignment Mismatch in Unsupervised Image Classification

Abstract

Unsupervised image classification is a challenging computer vision task. Deep learning-based algorithms have achieved superb results, where the latest approach adopts unified losses from embedding and class assignment processes. Since these processes inherently have different goals, jointly optimizing them may lead to a suboptimal solution. To address this limitation, we propose a novel two-stage algorithm in which an embedding module for pretraining precedes a refining module that concurrently performs embedding and class assignment. Our model outperforms SOTA when tested with multiple datasets, by substantially high accuracy of 81.0% for the CIFAR-10 dataset (i.e., increased by 19.3 percent points), 35.3% accuracy for CIFAR-100-20 (9.6 pp) and 66.5% accuracy for STL-10 (6.9 pp) in unsupervised tasks.

Cite

Text

Han et al. "Mitigating Embedding and Class Assignment Mismatch in Unsupervised Image Classification." Proceedings of the European Conference on Computer Vision (ECCV), 2020. doi:10.1007/978-3-030-58586-0_45

Markdown

[Han et al. "Mitigating Embedding and Class Assignment Mismatch in Unsupervised Image Classification." Proceedings of the European Conference on Computer Vision (ECCV), 2020.](https://mlanthology.org/eccv/2020/han2020eccv-mitigating/) doi:10.1007/978-3-030-58586-0_45

BibTeX

@inproceedings{han2020eccv-mitigating,
  title     = {{Mitigating Embedding and Class Assignment Mismatch in Unsupervised Image Classification}},
  author    = {Han, Sungwon and Park, Sungwon and Park, Sungkyu and Kim, Sundong and Cha, Meeyoung},
  booktitle = {Proceedings of the European Conference on Computer Vision (ECCV)},
  year      = {2020},
  doi       = {10.1007/978-3-030-58586-0_45},
  url       = {https://mlanthology.org/eccv/2020/han2020eccv-mitigating/}
}