The youtube video quotes a paper by the same author, so it's probably the same group's work. I wonder why didn't they used an approach similar to the one in the video using SD - it looks more viable.