Defects, such as dislocations, have a substantial adverse influence on the efficiency of semiconductor devices. Hence, characterising the type (i.e. edge or screw) and density of dislocations in semiconductor materials is a crucial technology for yield management and performance monitoring. Current technologies for holistic dislocation characterization use multiple time-consuming techniques, such as transmission electron microscopy, and have limited applicability for fast, in-line inspection.This project will change this, creating a new method of fast analysis of extended defects by training a machine learning (ML) model to identify dislocation density, type and properties (i.e. Burgers vector direction) from multidimensional cathodoluminescence (CL) data sets. It will take advantage of multi-microscopy experience, utilising the best techniques possible to train a ML model on CL datasets and then exploit the full information space provided by hyperspectral CL.