[brbsix]

If it was that easy to tell whether a course was pirated (yet alone determine the original owner), this whole mess probably wouldn't have happened. Udemy is in the best position to identify infringement since they have complete access to the courses, but it's anyone's guess whether they would go to such lengths. I would guess the best bet would take a few screens at random durations then perform an image search against a database of screens collected from courses on other marketplaces. Then use Mechanical Turk for final verification. I suppose you could perform facial recognition against profile photos of authors on other marketplaces as well (for the courses that actually include a headshot). I've used OpenBR to do similar things in the past but accuracy was too low to be useful in my case.