Online detection and classification of moving objects using progressively improving detectors

摘要: Boosting based detection methods have successfully been used for robust of faces and pedestrians. However, a very large amount labeled examples are required training such classifier. Moreover, once trained, the boosted classifier cannot adjust to particular scenario in which it is employed. In this paper, we propose co-training approach continuously label incoming data use online update that was initially trained from small example set. The main contribution our an procedure separate views (features) co-training, while combined view (all features) make classification decisions single framework. features derived principal component analysis appearance templates examples. order speed up classification, background modeling prune away stationary regions image. Our experiments indicate starting on set, significant performance gains can be made through updation unlabeled data.