Deep learning methods for designing AR display using VHOE-based waveguide

We propose a CNN model to simulate wave propagation in an AR optical system using Volume Holographic Optical Elements (VHOEs) as combiners. The architecture and CWT-based training dataset are discussed, focusing on chromatic dispersion-induced color distortion. After training, the model achieves nearly 100% color accuracy with unknown spectra, demonstrating deep learning’s potential for modeling complex optical systems and advancing HOE-based AR displays.