having an all-in-one solution that is easy to implement and works reliably would be VERY appealing to a lot of teams.
Honestly, I don’t think “works reliably” is enough. Personally, I’d more or less flat-out refuse to use anything that we can’t directly modify and edit the code on unless it works flawlessly 100% of the time. Especially if we’re spending as much — or more — money on the smart camera as we did our x86_64 coprocessor that had a desktop GPU.
While some teams do drop that amount of money on vision gear, for basic contour based vision, you honestly only need a $35 Raspberry Pi and a $7 Playstation 3 Eye camera (they’re purpose built for CV: 640x480@60fps or 320x240@150fps, shockingly cheap). The sort of team that would be willing to spend $400 on vision is — I’d bet in general — more inclined to implement it themselves unless this solution truly is flawless.
I think making this open source would certainly spark demand.