Human Detection via Classification on Riemannian Manifolds

We present a new algorithm to detect humans in still images utilizing covariance matrices as object descriptors. Since these descriptors do not lie on a vector space, well known machine learning techniques are not adequate to learn the classifiers. The space of d-dimensional nonsingular covariance matrices can be represented as a connected Riemannian manifold. We present a novel approach for classifying points lying on a Riemannian manifold by incorporating the a priori information about the geometry of the space. The algorithm is tested on INRIA human database where superior detection rates are observed over the previous approaches.

• Type: Conference
• Year: 2007
• Where: CVPR
• Authors: Oncel Tuzel, Fatih Porikli, Peter Meer
• Keywords: Computer Vision, Connected Riemannian Manifold, Covariance Matrices, CVPR 2007, Images Utilizing Covariance, Nonsingular Covariance Matrices, Object Descriptors