Training object detectors from few weakly-labeled and many unlabeled images

Zhaohui Yang, Miaojing Shi, Chao Xu, Vittorio Ferrari, Yannis Avrithis