Fig. 1: Conventional imaging and in-sensor computing architectures.

a Conventional imaging and postprocessing process. b In-sensor computing using a superconducting detector. The different color maps correspond to different convolution kernels. The size of the kernel is 5 × 5 × N, and N represents the number of multiplexed dimensions. Functions implemented by in-sensor computing include image classification, image preprocessing, and spectral classification. c Superconducting nanowire arrays with multiple programmable dimensions, including photon count rate, response time, pulse amplitude, and spectral responsivity. d Two different computing architectures, red represents PCR computing and blue represents area computing, and i represents the number of convolution kernels.