Getting High-Quality Data for Computer Vision Models