摘要
A physical neural network for THz imaging employs hybrid optical-digital signal dimension transformation to address hardware limitations, achieving high-quality, diffraction-free, real-time imaging at 2 frames per second with an expandable field of view.