Middle Output Regularized End-to-End Optimization for Computational Imaging

Published in Optica, 2023

We penalize divergences, variance, and structure of intermediate activations so that the learned single-pixel and CASSI encoders satisfy optical constraints and deliver better reconstructions.

Recommended citation: R Jacome, P Gomez, H Arguello. Middle output regularized end-to-end optimization for computational imaging. Optica 10 (11), 1421-1431, 2023.
Download Paper