A Comprehensive Survey of Image Augmentation Techniques for Deep Learning

Although deep learning has achieved satisfactory performance in computer vision, a large volume of images is required. However, collecting often expensive and challenging. Many image augmentation algorithms have been proposed to alleviate this issue. Understanding existing is, therefore, essential for finding suitable developing novel methods given task. In study, we perform comprehensive survey using informative taxonomy. To examine the basic objective augmentation, introduce challenges vision tasks vicinity distribution. The are then classified among three categories: model-free, model-based, optimizing policy-based. model-free category employs from processing, whereas model-based approach leverages generation models synthesize images. contrast, policy-based aims find an optimal combination operations. Based on analysis, believe that our enhances understanding necessary choosing designing algorithms.