Abstract
X-ray diffraction is a commonly used experimental science to detect the atomic and molecular structure of crystalline material. The process is called X-ray crystallography (XRC). Traditionally, it is done by human experts with some conjecture about what structure the crystalline material is likely to be. However, the study of crystal structure using X-ray diffraction patterns is applicable in many domains, such as chemistry, physics, biology, etc. It is tedious to have manual crystallography of X-ray diffraction patterns to determine a crystal structure with a massive amount of dataset. With the advent of high computational resources, deep learning techniques have taken classification to its peak. Convolution Neural Network (CNN) maps an input image into a high dimensional space and produce a low-cost function for image classification. In this paper, we deploy a variation of the Convolution Neural Network to predict crystal structure from X-ray diffraction patterns. We compare our approach with a wide range of conventional as well as modern Machine Learning based classification techniques for the structure prediction of a crystal. We report a cross-validation accuracy of 95.6% and Micro F1-score of 0.949 with our model for the proposed task which is significantly better than the other reported baseline methods.