It's exactly the stuff you see at 0:48 to 0:52 in the video. Computer vision using low resolution, 3d scale-invariant pattern recognition techniques for mass-public surveillance. Back then I was using SIFT. Nowadays, it must be neural net deep-learning techniques. Maybe not, since it's generally much slower and takes up plenty of resources compared to the lightning speed, low resolution SIFT feature gait-modelling and clustering techniques we applied back then. Felt kinda disturbed realising what we've made years after, but I guess it's just an inevitable piece of technology, which of course, could be used for both good and evil. Video by WSJ.
