Self-Supervised Learning by Estimating Twin Class Distribution
- 14 April 2023
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Image Processing
- Vol. 32 (10577149) , 2228-2236
Abstract
We present TWIST, a simple and theoretically explainable self-supervised representation learning method by classifying large-scale unlabeled datasets in an end-to-end way. We employ a siamese network terminated by a softmax operation to produce twin class distributions of two augmented images. Without supervision, we enforce the class distributions of different augmentations to be consistent. However, simply minimizing the divergence between augmentations will generate collapsed solutions, i.e., outputting the same class distribution for all images. In this case, little information about the input images is preserved. To solve this problem, we propose to maximize the mutual information between the input image and the output class predictions. Specifically, we minimize the entropy of the distribution for each sample to make the class prediction assertive, and maximize the entropy of the mean distribution to make the predictions of different samples diverse. In this way, TWIST can naturally avoid the collapsed solutions without specific designs such as asymmetric network, stop-gradient operation, or momentum encoder. As a result, TWIST outperforms previous state-of-the-art methods on a wide range of tasks. Specifically on the semi-supervised classification task, TWIST achieves 61.2% top-1 accuracy with 1% ImageNet labels using a ResNet-50 as backbone, surpassing previous best results by an improvement of 6.2%. Codes and pre-trained models are available at https://github.com/bytedance/TWIST.Keywords
All Related Versions
Funding Information
- National Natural Science Fund for Key International Collaboration (62120106005)
This publication has 37 references indexed in Scilit:
- Unsupervised Visual Representation Learning by Context Prediction2015
- Going deeper with convolutions2015
- Fully convolutional networks for semantic segmentation2015
- 3D Object Representations for Fine-Grained Categorization2013
- Cats and dogs2012
- The Pascal Visual Object Classes (VOC) ChallengeInternational Journal of Computer Vision, 2009
- Learning Generative Visual Models from Few Training Examples: An Incremental Bayesian Approach Tested on 101 Object Categories2005
- The Hungarian method for the assignment problemNaval Research Logistics Quarterly, 1955
- On Information and SufficiencyThe Annals of Mathematical Statistics, 1951
- A Mathematical Theory of CommunicationBell System Technical Journal, 1948