So it might thwart some newer learning algos, but I wonder how it would do against the more classical heuristically guided techniques that tend to look at more

So it might thwart some newer learning algos, but I wonder how it would do against the more classical heuristically guided techniques that tend to look at more invariant characteristics like facial feature spacing and ratios.