ContraCluster: Learning to Classify without Labels by Contrastive Self-Supervision and Prototype-Based Semi-Supervision

• Published in: arXiv.org
• Main Authors: Seongho Joe, Kim, Byoungjip, Kang, Hoyoung, Park, Kyoungwon, Bogun Kim, Park, Jaeseon, Lee, Joonseok, Gwon, Youngjune
• Format: Paper Journal Article
• Language: English
• Published: Ithaca Cornell University Library, arXiv.org 19.04.2023
• Subjects: Accuracy; Clustering; Computer Science - Computer Vision and Pattern Recognition; Computer Science - Learning; Image classification; Labels; Prototypes; Self-supervised learning
• ISSN: 2331-8422
• DOI: 10.48550/arxiv.2304.09369

The recent advances in representation learning inspire us to take on the challenging problem of unsupervised image classification tasks in a principled way. We propose ContraCluster, an unsupervised image classification method that combines clustering with the power of contrastive self-supervised learning. ContraCluster consists of three stages: (1) contrastive self-supervised pre-training (CPT), (2) contrastive prototype sampling (CPS), and (3) prototype-based semi-supervised fine-tuning (PB-SFT). CPS can select highly accurate, categorically prototypical images in an embedding space learned by contrastive learning. We use sampled prototypes as noisy labeled data to perform semi-supervised fine-tuning (PB-SFT), leveraging small prototypes and large unlabeled data to further enhance the accuracy. We demonstrate empirically that ContraCluster achieves new state-of-the-art results for standard benchmark datasets including CIFAR-10, STL-10, and ImageNet-10. For example, ContraCluster achieves about 90.8% accuracy for CIFAR-10, which outperforms DAC (52.2%), IIC (61.7%), and SCAN (87.6%) by a large margin. Without any labels, ContraCluster can achieve a 90.8% accuracy that is comparable to 95.8% by the best supervised counterpart.