On 3D object retrieval benchmarking

AbstractThe continuous evolution of 3D computer graphics and the progress of 3D digitization systems resulted in a continuous increase in the available 3D content. The widespread use of 3D objects in diverse domains contributed on forming 3D object retrieval as an active research field. In order to objectively evaluate the performance of retrieval methodologies there is a need for objective benchmarking schemes. In this work, we provide a comprehensive overview of the state-of-the-art evaluation methodologies including not only the performance measures but also the corresponding benchmark datasets. Meaningful benchmark datasets are discussed while a detailed list of publicly available 3D model repositories is given organized in terms of application domains, content magnitude and data types.