From the paper: “ To the best of our knowledge, we are the first to demonstrate deep learning data-driven 3D photorealistic reconstruction without system calib

From the paper: “ To the best of our knowledge, we are the first to demonstrate deep learning data-driven 3D photorealistic reconstruction without system calibration and initializations, and we are the first to demonstrate imaging objects behind obstacles using lensless imagers.” That’s their actual claim, worth examining. Also AFAICT their camera has no large traditional lens at all, whereas previous work — ie lytro — put a traditional lens in front of their micro lens array.