Boosting Object Detection in Cyberphysical Systems

The construction of Cyberphysical systems requires providing intelligent behavior to physical agents at the smallest scale and, therefore, the need to develop very efficient and resource-aware algorithms. In this paper we present an object detection algorithm that may endow an agent with perceptual object detection capabilities at a small computational cost. To this end we adapt a recent Multi-class Boosting scheme to create an efficient detector with the capability of regressing the object bounding box. In the experiments we prove that the resulting algorithm shows Average Precision (AP) improvements in a multi-view car detection problem.