In this thesis project you will be working with Veoneer in Linköping, in collaboration with the division for Media and Information Technology at Campus Norrköping. The focus is on exploring the possibilities of improving machine learning applications, and in particular deep learning, using synthetic image data. It is difficult, expensive and time-consuming to capture all possible traffic situations in a training dataset of real images. For example, accidents are rare occasions, but very important to manage correctly in an autonomous application. Thus, images that have been generated by means of computer graphics can prove to be a valuable source for improving the end performance of a deep learning application in autonomous driving. The project will look closer on how to incorporate synthetic images in training, how to properly transfer the result to real images, and how to test the impact on the targeted application, e.g. for semantic segmentation.
Contact: Gabriel Eilertsen (firstname.lastname@example.org), division of Media and Information Technology, Department of Science and Technology, Campus Norrköping, Linköping University.