| Abstract: |
Recently, content based 3D shape retrieval has drawn more and more attention by researchers. How to properly evaluate the result of shape searching algorithms is important. Benchmarking is one answer to this question. Due to the limitations of current shape benchmarks, in this paper, we propose a new publicly available 3D shape benchmark to advance the state of art in 3D shape retrieval. A thorough description of the new shape benchmark is also given. First, the 3D models are classified mainly according to the shape similarity; second, the geometric structure of each model is modified and normalized, and each class in the benchmark shares the equal number of models to reduce the possible bias of evaluation results; finally,several representative algorithms are evaluated on the benchmark,and an experiment is also conducted to show the reliability of the new benchmark. |
Click here to retrieve PDF version of paper (2MB)