We want to reduce exposure time of camera, but that would mean lesser time for collection of photons, resulting in a dimmer image.
We would like to train a GAN on pairs of {high exposure => ground truth, low exposure => input}.
This should help us reduce exposure time globally, and improve system latency.
Also, we reduce the destructive experimental burden on any given cell by reducing photon exposure.