Car/Non-Car Classification in an Informative Sample Subspace

In this paper, we present a method for data classification with application to car/non-car objects. We first developed a sample based car/non-car maximal mutual information low dimensional subspace. We then trained a support vector machine (SVM) in this subspace for the detection of cars. Using publicly available standard training and testing data sets, we demonstrated that our car detector gave very competitive performances.

• Type: Conference
• Year: 2006
• Where: ICPR
• Authors: Guoping Qiu, Jianzhong Fang
• Keywords: Computer Vision, ICPR 2006, Low Dimensional Subspace, Maximal Mutual Information, Support Vector Machine