Abstract
Unsupervised deep learning has only been used in rotationally symmetric optical design. This work presents a differentiable three-dimensional ray tracing module and related loss functions, enabling unsupervised learning of non-rationally symmetric freeform optical systems.
© 2023 SPIE
PDF ArticleMore Like This
Congli Wang
FM3A.2 Frontiers in Optics (FiO) 2023
Xinge Yang, Qiang Fu, and Wolfgang Heidrich
CTh4C.2 Computational Optical Sensing and Imaging (COSI) 2022
Dushan N. Wadduwage and Udith Haputhanthri
JTu4A.43 3D Image Acquisition and Display: Technology, Perception and Applications (3D) 2023