Maurice Peemen (Thermo Fisher Scientific)

Image quality evaluation by deep learning enabled segmentation

Of the many scientific lab products made by Thermo Fisher Scientific the electron microscopes are among the most complex and advanced. These tools are able to image subnanometer structures in complex materials for various life science, material science and semiconductor use cases. For these use cases the information in images is of key importance. To objectively evaluate the information content and image quality classical methods like the PSNR or SSIM scores are often too simplistic. They prefer smooth images while our images are often dominated by noise. Deep learning has brought major improvements in image segmentation techniques, enabling us to analyze and evaluate the information content that is required for end user applications.

Maurice Peemen received a bachelor’s degree in electrical engineering from Fontys University of Applied Sciences in 2007. His graduation project was at CNSE in Albany, New York, USA on an EUV lithography mask flatness metrology tool. In 2010 he obtained his MSc degree in electrical engineering with the predicate ‘cum laude’ from Eindhoven University of Technology. Later in 2011 he started as a PhD student under the supervision of Henk Corporaal in the Electronic Systems group at the TUE. His dissertation work was on improving the efficiency of deep convolutional networks. This work contained topics in deep learning, loop transformations and accelerator architectures. In 2015 he left the ES group to start as a scientist at Thermo Fisher Scientific (formerly FEI Company) to work on high-performance microscopy workflow solutions.