In this whitepaper, we concentrated on active learning algorithms, which help deep learning engineers select a subset of images from a large unlabeled pool of data in such a way, that obtaining annotations of those images will result in a maximal increase of model accuracy.
