[2000 CVPR] A Statistical Method for 3D Object Detection Applied to Faces and Cars英文版本.pdf

A Statistical Method for 3D Object Detection Applied to Faces and Cars Henry Schneiderman and Takeo Kanade Robotics Institute Carnegie Mellon University Pittsburgh, PA 15213 Abstract is specialized to frontal views. We apply these view-based In this paper, we describe a statistical method for 3D object detectors in parallel and then combine their results. If there are detection. We represent the statistics of both object appear- multiple detections at the same or adjacent locations, our ance and “non-object” appearance using a product of histo- method chooses the strongest detection. grams. Each histogram represents the joint statistics of a subset We empirically determined the number of orientations to of wavelet coefficients and their position on the object. Our model for each object. For faces we use two view-based detec- approach is to use many such histograms representing a wide tors: frontal and right profile, as shown below. To detect left- variety of visual attributes. Using this method, we have devel- profile faces, we apply the right profile detector to a mirror- oped the first algorithm that can reliably detect human faces reversed input images. For cars we use eight detectors as with out-of-plane rotation and the first algorithm that can reli- shown below. Again, we detect left side views by running the ably detect passenger c