EnAET: A Self-Trained Framework for Semi-Supervised and Supervised Learning With Ensemble Transformations

Deep neural networks have been successfully applied to many real-world applications. However, such successes rely heavily on large amounts of labeled data that is expensive obtain. Recently, methods for semi-supervised learning proposed and achieved excellent performance. In this study, we propose a new EnAET framework further improve existing with self-supervised information. To our best knowledge, all current performance prediction consistency confidence ideas. We are the first explore role representations in under rich family transformations. Consequently, can integrate information as regularization term methods. experiments, use MixMatch, which state-of-the-art method learning, baseline test framework. Across different datasets, adopt same hyper-parameters, greatly improves generalization ability Experiment results datasets demonstrate algorithms. Moreover, also supervised by margin, including extremely challenging scenarios only 10 images per class. The code experiment records available https://github.com/maple-research-lab/EnAET.