Inpainting microstructure with GANs

Imaging is critical to the characterisation of materials. However, imaging and sample preparation techniques are often sensitive to defects and unwanted artefacts. Imaging and sample preparation can be expensive and time consuming and for many applications, the whole sample is required for simulation or analysis to obtain representative results. Microstructural inpainting is a method to alleviate this problem by replacing occluded regions with generated microstructure. In this paper we introduce two methods using generative adversarial networks to generate contiguous inpainted regions of arbitrary shape and size by learning the microstructural distribution from the unoccluded data. We also outline the development of a graphical user interface that allows users to utilise these machine learning methods in a `no-code' environment.

Issac Squires is a PhD student in the tldr group in the Dyson School of Design Engineering. He uses machine learning and software development to build tools that aid the design and optimisation of next generation battery materials.

Energy Futures Lab is one of seven Global Institutes at Imperial College London. The institute was established to address global energy challenges by identifying and leading new opportunities to serve industry, government and society at large through high quality research, evidence and advocacy for positive change. The institute aims to promote energy innovation and advance systemic solutions for a sustainable energy future by bringing together the science, engineering and policy expertise at Imperial and fostering collaboration with a wide variety of external partners.

Event Venue & Nearby Stays

Room 611, Electrical & Electronic Engineering Building, Imperial College London, London, United Kingdom