TY - GEN
T1 - Nonrigid 3D shape retrieval with HAPPS
T2 - 6th Annual International Conference on Computational Science and Computational Intelligence, CSCI 2019
AU - Otu, Ekpo
AU - Zwiggelaar, Reyer
AU - Hunter, David
AU - Liu, Yonghuai
PY - 2020/4/20
Y1 - 2020/4/20
N2 - A robust, yet computationally efficient signature for describing 3D shape remains a challenge for 3D computer vision and related applications. Having a signature that is generalizable across a wider range of datasets becomes another important research issue. This paper proposes a novel Hybrid signature, the Augmented Point Pair Signature (HAPPS), that is robust, highly discriminating, efficient, and capable of effectively representing 3D point cloud and polygon mesh surfaces. We tested the overall performances of HAPPS on three standardized benchmark datasets for 3D shape retrieval: The Shape Retrieval Contest 2018 (SHREC'18) protein shapes benchmark, with 2,267 protein conformers, SHREC'17 Point cloud Retrieval of Nonrigid Toys (PRoNTo), with 100 3D point clouds, and SHREC'10 Nonrigid shape retrieval having 200 triangular meshes. Using 6 standard retrieval performance metrics to evaluate our results, we demonstrated the superiority of our HAPPS retrieval method over several other state-of-the-art methods for the SHREC'18 protein dataset, while also competing side-by-side with the best 2 performing methods for the other benchmark datasets.
AB - A robust, yet computationally efficient signature for describing 3D shape remains a challenge for 3D computer vision and related applications. Having a signature that is generalizable across a wider range of datasets becomes another important research issue. This paper proposes a novel Hybrid signature, the Augmented Point Pair Signature (HAPPS), that is robust, highly discriminating, efficient, and capable of effectively representing 3D point cloud and polygon mesh surfaces. We tested the overall performances of HAPPS on three standardized benchmark datasets for 3D shape retrieval: The Shape Retrieval Contest 2018 (SHREC'18) protein shapes benchmark, with 2,267 protein conformers, SHREC'17 Point cloud Retrieval of Nonrigid Toys (PRoNTo), with 100 3D point clouds, and SHREC'10 Nonrigid shape retrieval having 200 triangular meshes. Using 6 standard retrieval performance metrics to evaluate our results, we demonstrated the superiority of our HAPPS retrieval method over several other state-of-the-art methods for the SHREC'18 protein dataset, while also competing side-by-side with the best 2 performing methods for the other benchmark datasets.
KW - 3D shape analysis
KW - Content-based retrieval
KW - Local/Global descriptor
KW - Shapes descriptors
UR - http://www.scopus.com/inward/record.url?scp=85084747374&partnerID=8YFLogxK
U2 - 10.1109/CSCI49370.2019.00124
DO - 10.1109/CSCI49370.2019.00124
M3 - Conference Proceeding (Non-Journal item)
AN - SCOPUS:85084747374
SN - 9781728155852
T3 - Proceedings - 6th Annual Conference on Computational Science and Computational Intelligence, CSCI 2019
SP - 662
EP - 668
BT - Proceedings - 6th Annual Conference on Computational Science and Computational Intelligence, CSCI 2019
PB - IEEE Press
Y2 - 5 December 2019 through 7 December 2019
ER -