[frontierkodiak]

That's the main approach I've encountered. Slice, inference, filter duplicates from the results.

https://github.com/obss/sahi