Auto Data Augmentation for Image: A Brief Survey
Xuan Xia, Jingfei Zhang, Xing He, Haoran Tong, Xiaoguang Zhang, Nan Li, Ning Ding · 2023
Auto data augmentation has emerged as a promising alternative to the laborious manual parameter tuning involved in data augmentation policies. However, the existing approaches have limitations in terms of their applicability to a restricted range of models, datasets, and tasks. In this brief survey, we identify and discuss four key issues that auto data augmentation needs to tackle. We categorize auto data augmentation into two main types: closed-loop/open-loop auto data augmentation and online/offline auto data augmentation. Each of these categories is examined in detail, with a comprehensive analysis of their respective performance. Furthermore, we highlight the challenges that the field of auto data augmentation faces and offer insights into potential research directions. Our survey serves as a valuable resource for researchers seeking a deeper understanding of the evolving landscape of auto data augmentation, providing inspiration and guidance for future work in this area.