Meta-Optics for Image and Spectral Classification

Here, we demonstrate how meta-optics can be designed, and realized, to work in concert with a digital back-end for a range of tasks including image and spectral classification. In the case of image classification, the meta-optic serves to off-load computationally expensive tasks into an optical front-end, speeding computational speed while lowering energy consumption. For spectral classification, meta-optics serve as complex filters which are used in conjunction with a digital neural network to achieve a compact flow cytometer capable of classifying up to four unique fluorophores, and their combinations.