Optical neural networks in the quantum domain

Le résumé fourni par la source

From self-driving cars to medical diagnostics, computer vision underpins technologies that identify objects, recognize patterns, and analyze images. We considers computer vision in situations when only faint light is available, such as reading a road sign at night, examining fragile biological cells under a microscope, or detecting distant exoplanets. In these settings, its performance is degraded due to quantum nature of light – specifically the shot noise associated with random arrival of individual photons. We show how these limits can be overcome by marrying computer vision with quantum physics. The central idea is to treat vision tasks as quantum sensing problems. Instead of image capture followed by digital analysis, we to design and train optical neural networks to extract information in the most efficient way permitted by quantum mechanics.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.