Simon Proud

Hot take: deep learning is slowly going to make large swaths of remote sensing research irrelevant. It takes a PhD to understand sensors and physics and ecology enough to write an algorithm to deduce crop type. It takes a weekend to train a deep learning model to do it.

Great! So how come after over a decade of machine learning research to produce cloud masks for remote sensing days there's still very few that perform better than the expert cloudmasks? That's a lot of weekends of training we've had!

days = data. Machine learning text prediction ain't as good as expert typing either ;-)