Curtaining artifacts generation on synthetic FIB-SEM data via Generative Adversarial Networks

Abstract

FIB-SEM imaging stands out as an advanced method for capturing nanoscale structures. Throughout the image acquisition process, various artifacts emerge, including curtaining and charging artifacts. Effectively addressing these artifacts requires specialized algorithms tailored to their unique characteristics. Consequently, the development of algorithms demands simulated images used as benchmarks for validation. Simulating FIB-SEM images is a complex task, prompting the exploration of generative models as an alternative for simulation. We have adapted generative models to encompass curtaining artifacts, a feature challenging to replicate through conventional simulations. The resulting images demonstrate comparability with synthetically generated counterparts.