Progress in applying supervised learning for nat- ural scene classification is impeded by the lack of appropriate datasets for training. This paper describes the automatic generation of synthetic three-dimensional (3D) ...