Mask defect detection with hybrid deep learning network

Background: Deep learning is a very fast-growing field in the area of artificial intelligence with remarkable results recent years. Many works lithography and photomask have shown progress technology this application potential to improve by automation processes. Despite progress, use machine techniques mask repair still seems be at beginning. Aim: We show that deep learning-based methods can successfully applied applications. Approach: The presented system hybrid modular approach based on combination several networks analytical methods, enabling detection pattern defects determination exact defect shapes from SEM images. In current version, trained for line/space patterns contact typical types. modularity allows extensibility new cases. issue an insufficient amount training data addressed using purely computer-generated simplified specific network architecture. Results: good functionality accuracy are demonstrated set real images numerous defects. particular, 100% true rate could obtained. Conclusions: demonstrates successful identification, location, shape