What Neural Networks can do for Intelligence, Surveillance, and Reconnaissance

Dahl, R., Norouzi, M., & Shlens, J. (2017). probabilistic pixel recursive super resolution model

According to this study by the Google Brain team "At increasing levels of magnification, the details do not exist in the source image anymore, and the predictive challenge shifts from recovering details (e.g., deconvolution [23]) to synthesizing plausible novel details de novo [33, 44]." (Dahl, R., Norouzi, M., & Shlens, J. 2017) With the ability for machine learning to fill in the gaps, it's possible to take an image from far away and utilize these systems to create a less pixelated and more defined image to use for intelligence.

This type of technology would never replace a better camera or a better-positioned system; however, there may be situations where you have little time to use the best camera available. This type of technology would revolutionize how images are taken and how they are processed for the best information possible.

In the Intelligence Surveillance and Reconnaissance world, there can be times when you can't always capture the image that you want. This can be due to many factors, including weather and distance between your target and your platform. With machine learning and algorithms designed to complete images, this could be a problem of the past.

Dahl, R., Norouzi, M., & Shlens, J. (2017). Pixel recursive super-resolution. In Proceedings of the IEEE International Conference on Computer Vision (pp. 5439-5448). https://arxiv.org/pdf/1702.00783.pdf?xtor=AL-32280680