Deep Learning Approaches to Image Texture Analysis in Material Processing

Texture analysis is key to better understanding of the relationships between microstructures materials and their properties, as well use models in process systems using raw signals or images input. Recently, new methods based on transfer learning with deep neural networks have become established highly competitive approaches classical texture analysis. In this study, three traditional approaches, grey level co-occurrence matrices, local binary patterns textons are compared five AlexNet, VGG19, ResNet50, GoogLeNet MobileNetV2. This done two simulated one real-world case study. studies, material were Voronoi graphic representations appearance ultrahigh carbon steel cast a textural pattern recognition pattern. The ability random forest models, convolutional themselves, discriminate different textures image features input was used basis for comparison. texton algorithm performed than LBP GLCM algorithms similar when these directly, without any retraining. Partial full retraining yielded considerably results, MobileNetV2 yielding best results.