Alumina spatial hetereogeneity quantification assisted by deep learning

This module provides deep learning semantic segmentation propoped in [1] for the quantification of spatial hetereogeneity of Gamma Alumina from SEM images. The architecture is a standard Unet encoder decoder [2], with a supervised training and inference using stochastic patches procedure [3]. In [1] results is then segmented in binary image with automatic histogram segmentation and post processed with several modules provided bellow. Some sample images are also provided.