Image Data Augmentation Approaches: A Comprehensive Survey and Future directions (original) (raw)

2023, arXiv (Cornell University)

Deep learning algorithms have demonstrated remarkable performance in various computer vision tasks, however, limited labeled data can lead to overfitting problems, hindering the network's performance on unseen data. To address this issue, various generalization techniques have been proposed, including dropout, normalization, and advanced data augmentation. Among these techniques, image data augmentationwhich increases the dataset size by incorporating sample diversity-has received significant attention in recent times. In this survey, we focus on advanced image data augmentation techniques. We provide an overview of data augmentation, present a novel and comprehensive taxonomy of the reviewed data augmentation techniques, and discuss their strengths and limitations. Furthermore, we provide comprehensive results of the impact of data augmentation on three popular computer vision tasks: image classification, object detection, and semantic segmentation. For results reproducibility, the available codes of all data augmentation techniques have been compiled. Finally, we discuss the challenges and difficulties, as well as possible future directions for the research community. This survey provides several benefits: i) readers will gain a deeper understanding of how data augmentation can help address overfitting problems, ii) researchers will save time searching for comparison results, iii) the codes for the data augmentation techniques are available for result reproducibility, and iv) the discussion of future work will spark interest in the research community.