[retrocryptid]

Having written algorithms for CV and satellite image processing, I can tell you that advancement in the field more or less halted in 2018 when everyone decided to dump money and time into CNNs.

I wrote a cloud detection algorithm that would run well on a 20MHz microcontroller powered by 4 AA batteries. But my bosses were worried they couldn't find devs to replace me if I retired, and figured a CS intern and a trained dog [*] could train a CNN to do the same thing. $600k and 2 years later they got something that worked until the dog got a better job and the intern returned to school. None of the surviving engineers could make heads or tails of the solution so they hired a PhD to replicate the work. After publishing his research paper he left and found a job at an east coast university.

So... Sure... CNNs are doing all sorts of things, mostly things we did with DFTs in the past and not quite as well.

* the joke here is the dog does the technical tasks while the intern's only responsibilities are to feed and walk the dog and give it a treat when it completes it's user stories before the end of the sprint.*

[moomoo11]

I feel like a cat after reading this.