Selective Aerial Image Labeling Matches Full Supervised Accuracy Using Six Percent Labels
A semi-supervised framework combining unsupervised feature learning with hybrid instance selection matched fully supervised aerial wildlife classification using just 6.4 ...

