Review of Image Augmentation Used in Deep Learning-Based Material Microscopic Image Segmentation

The deep learning-based image segmentation approach has evolved into the mainstream of target detection and shape characterization in microscopic analysis. However, accuracy generalizability learning approaches are still hindered by insufficient data problem that results from high expense human material resources for acquisition annotation. Generally, augmentation can increase amount a short time means mathematical simulation, become necessary module In this work, we first review commonly used methods divide more than 60 basic eleven categories based on different implementation strategies. Secondly, conduct experiments to verify effectiveness various task two classical images using evaluation metrics with applicabilities. U-Net model was selected as representative benchmark tasks, it is classic most widely field. We utilize improvement performance methods. Then, discuss advantages applicability task. conclusions work serve guide creation intelligent modeling frameworks materials industry.