A large-scale Shape Benchmark for 3D object retrieval: Toyohashi shape benchmark

In this paper, we describe the Toyohashi Shape Benchmark (TSB), a publicly available new database of polygonal models collected from the World Wide Web, consisting of 10,000 models, as the largest 3D shape models to our knowledge used for benchmark testing. TSB includes 352 categories with labels. It can be used for both 3D shape retrieval and 3D shape classification. Until now, the most well-known 3D shape benchmark has been the PSB, or the Princeton Shape Benchmark, consisting of 1,814 models, including the half as training data and the remaining half as testing. The TSB is approximately 6 times larger than the PSB. Unlike textual data such as TREC and NTCIR data collections, 3D shape repositories have been suffering from the shortage of data, and from the difficulty in testing the scalability of any algorithms that work on top of given benchmark data set. In addition to the TSB, we propose a new shape descriptor which we call DB-VLAT (Depth-Buffered Vector of Locally Aggregated Tensors). During the comparison with the TSB, we will demonstrate that our new shape descriptor exhibits the best search performance among those known programs to which we have had access on the Internet, including the Spherical Harmonic Descriptor and Light-Field Descriptor. We consider that the TSB can be a step toward the next generation 3D shape benchmark having massive 3D data collection, and hope it will serve for many purposes in both academia and industry.