Companies are using advanced neural network techniques for compression in static images and videos, based on image and video statistics, with success comparable to image compression starting from raw pixels to identifying objects.