DEEP LEARNING APPROACH BASED OPTICAL EDGE DETECTION USING ENZ LAYERS (INVITED)

Metamaterials offer a chance to design films that could achieve optical differentiation due to their special properties. Layered film would be the simplest case considering the easy-fabrication and compactness. Instead of performing the optical differentiation at the Fourier plane, Green-function based multi-layers are used to achieve optical differentiation. In this work, epsilon-near-zero (ENZ) material is utilized to realize the optical differentiation owning to the special optical properties that the reflection increases with the increase of incident angle, which fits the characteristics of optical differentiation. In addition, deep learning is also used in this work to simplify the design of ENZ layers to achieve the optical differentiation, and further realize the optical edge detection. Simulations based on the Fresnel diffraction are carried out to verify that our films designed by this method could realize the optical detection under different cases.