meh, he doesn't really address the issue. He just mentions using OBS to threshold the image and dilate the resulting regions.
That raises a lot of other questions: He obviously feeds two different videos to the displays. How is this done? Does it add latency, does it work with, let's say, Netflix DRM? Are the two videos even in sync or is the obs post processed feed always N frames behind (think of the transition between barely lit and bright scenes)?