Task Augmentation by Rotating for Meta-Learning

TL;DR

This paper proposes a task augmentation method by rotating images to increase class diversity, improving meta-learning performance on few-shot benchmarks by reducing overfitting.

Contribution

It introduces a novel task augmentation technique by rotating images to enhance class diversity for meta-learning, outperforming traditional image augmentation methods.

Findings

Achieves state-of-the-art results on miniImageNet, CIFAR-FS, and FC100.

Improves meta-learning training stability with less overfitting.

Outperforms rotation-based image augmentation methods.

Abstract

Data augmentation is one of the most effective approaches for improving the accuracy of modern machine learning models, and it is also indispensable to train a deep model for meta-learning. In this paper, we introduce a task augmentation method by rotating, which increases the number of classes by rotating the original images 90, 180 and 270 degrees, different from traditional augmentation methods which increase the number of images. With a larger amount of classes, we can sample more diverse task instances during training. Therefore, task augmentation by rotating allows us to train a deep network by meta-learning methods with little over-fitting. Experimental results show that our approach is better than the rotation for increasing the number of images and achieves state-of-the-art performance on miniImageNet, CIFAR-FS, and FC100 few-shot learning benchmarks. The code is available on \url{www.github.com/AceChuse/TaskLevelAug}.