I worked a NASA grant with a few others, combined, we had about 75 years of experience in our fields between the 3 of us. We demonstrated onboard computer vision labeling of geographic features of Jovian ice moons using an emulated BAE RAD750, and a very minimal neural network trained on existing Gallileo imaging data. We were very close to getting into talks to be a Tech Demo on Europa Clipper. Then, a random NASA program manager shut it all down because "I just know, you cant run a CNN labeling on a RAD750," despite the fact that we had easily reproducible hard evidence that it was possible.
That was the day I lost a lot of faith in NASA, they are not the organization they used to be.
