Unsupervised improvement of visual detectors using cotraining

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 is described, based on the cotraining framework of Blum and Mitchell.

Abstract

• Traffic camera taking video • You want to identify the cars • Simple • pick a detector • gather training data • train Figure 1: Example images used to test and train the car detection system. On the left are the original images. On the right are background subtracted images. Figure 2: Left: A scatter plot of the joint distribution of margins for the two classifiers. These results are shown on test data, and therefore represents the distribution on unlabeled data (positive examples are circles, negative are grey/green). For each classifier two threshold are also shown, the threshold above which no negative is found, , and the threshold below which no positive is found,. The regions labeled A,B,C, and D contain informative examples. Right: Particular examples taken from A, B, C, or D; images which are mislabeled by one classifier (or have small margin) which are confidently labeled by the other classifier. E.G. Set B contains images confidently labeled