3D Deep Shape Descriptor

Yi Fang, Jin Xie, Guoxian Dai, Meng Wang, Fan Zhu, Tiantian Xu, Edward Wong; The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2015, pp. 2319-2328

Abstract


Shape descriptor is a concise yet informative representation that provides a 3D object with an identification as a member of some category. This paper developed a concise deep shape descriptor for the first time to address challenging issues from ever-growing 3D datasets in areas as diverse as engineering, medicine, and biology. Specifically, the proposed approach developed novel techniques to extract concise but geometrically informative shape descriptor, new definitions of Eigen-shape descriptor and Fisher-shape descriptor to guide the training strategy for deep neural network, and deep shape descriptor with discriminative capacity of maximizing the inter-class margin while minimizing the intra-class variance. Our approach addressed the challenges for shape analysis techniques posed by the complexity of 3D model and data representation and geometric structural variations and noise present in 3D models. The experimental results on 3D shape retrieval demonstrate that our proposed deep shape descriptor is superior to other state-of-the-art approaches on handling noise, incompleteness and 3D shape structural variations.

Related Material


[pdf]
[bibtex]
@InProceedings{Fang_2015_CVPR,
author = {Fang, Yi and Xie, Jin and Dai, Guoxian and Wang, Meng and Zhu, Fan and Xu, Tiantian and Wong, Edward},
title = {3D Deep Shape Descriptor},
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2015}
}