Using atmospheric understanding to exploit multi-spectral images

Ankit Patnala published his first paper in the IEEE journal.
There is yet more information currently unused by machine learning in satellite data. Natural images come in RGB and thus most machine learning models working with image data, process RGB channels, ignoring valuable information carried in other channels. In this publication, Patnala et al. show that the near-infrared channel has additional information which helps the machine learning model learn.

Check out the early access paper here: