Using deep learning-based training of diffractive layers we designed single-pixel machine vision systems to all-optically classify images by maximizing the output power of the wavelength corresponding to the correct data-class. We experimentally validated our diffractive designs using a plasmonic nanoantenna-based time-domain spectroscopy setup and 3D-printed diffractive layers to successfully classify the images of handwritten-digits using a single-pixel and snap-shot illumination. Furthermore, we trained a shallow electronic neural network as a decoder to reconstruct the images of the input objects, solely from the power detected at ten distinct wavelengths, also demonstrating the success of this platform as a task-specific, single-pixel imager.
Most state-of-the-art terahertz time-domain imaging technologies are based on single-pixel systems, which mechanically scan either the imaging object or the terahertz system, limiting the imaging speed. We present a new terahertz time-domain imaging modality using a terahertz photoconductive focal-plane array. The focal-plane array consists of plasmonic nano-antenna arrays on an LT-GaAs substrate. The dynamic range of a single pixel can reach up to 75 dB with more than a 4 THz bandwidth. We demonstrate clear terahertz images up to 2.5 THz. We also demonstrate that the focal-plane array can operate at video-rate imaging speeds.