Unsupervised improvement of visual detectors using cotraining

Abstract

One significant challenge in the construction of visual detection systems is the acquisition of sufficient labeled data. This paper describes a new technique for training visual detectors which requires only a small quantity of labeled data, and then uses unlabeled data to improve performance over time. Unsupervised improvement is based on the cotraining framework of Blum and Mitchell, in which two disparate classifiers are trained simultaneously. Unlabeled examples which are confidently labeled by one classifier are added, with labels, to the training set of the other classifier. Experiments are presented on the realistic task of automobile detection in roadway surveillance video. In this application, co-training reduces the false positive rate by a factor of 2 to 11 from the classifier trained with labeled data alone.

Automatic adaptation of a generic pedestrian detector to a specific traffic scene

“…They controls how aggressive the automatic training process is. Similar parameters also exist in other approaches of automatically training scene specific detectors [10,11,12,13,14,15,16]. Our approach has robustness to these parameters within certain range.…”

“…By calculating the frame difference as 0.5(|I t − I t−50 | + I t − I t+50 |), moving pixels inside a detection window are thresholded and counted. Similar to other self-training [15,13] or co-training [11,10,16,21] frameworks, the confident positive examples are found by thresholding L p (z) > L 0 . The larger the threshold is, the more conservative the strategy of selecting examples is.…”

“…The prediction of one detector on unlabeled examples was used to enlarge the training set of the other. For example, Levin et al [11] built two car detectors using gray images and background subtracted images. They all required manually labeling a small training set from the target scene for initialization.…”

“…Some automatic labelers were designed under the cotraining framework [11,10,16,21]. Two detectors based on different types of features were trained iteratively.…”

Adapting pedestrian detectors to new domains: A comprehensive review

“…Levin et al [59] propose a co-training [60] approach for semi-supervised learning of vehicle detectors in video. Given some labelled data in the target scene, firstly, a pair of car detectors is trained; one of the pairs is trained on data for whose feature extraction is performed on original images and for the other, background subtracted images instead of the original images are used.…”