An Evaluation of Canonical Forms for Non-rigid 3D Shape Retrieval

David Pickup
Cardiff University
Juncheng Liu
Cardiff University
Xianfang Sun
Cardiff University
Paul L. Rosin
Cardiff University

Ralph R. Martin
Cardiff University
Zhiquan Cheng
Avatar Science
Zhouhui Lian
Peking University
Sipin Nie
South China University of Technology

Longcun Jin
South China University of Technology
Gil Shamai
Yusuf Sahillioglu
Middle East Technical University
Ladislav Kavan
University of Utah

Selection of models from the Real Human dataset.


Canonical forms attempt to factor out a non-rigid shape's pose, giving a pose-neutral shape. This opens up the possibility of using methods originally designed for rigid shape retrieval for the task of non-rigid shape retrieval. We extend our recent benchmark for testing canonical form algorithms. Our new benchmark is used to evaluate a greater number of state-of-the-art canonical forms, on five recent non-rigid retrieval datasets, within two different retrieval frameworks. A total of fifteen different canonical form methods are compared. We find that the difference in retrieval accuracy between different canonical form methods is small, but varies significantly across different datasets. We also find that efficiency is the main difference between the methods.


David Pickup, Juncheng Liu, Xianfang Sun, Paul L. Rosin, Ralph R. Martin, Zhiquan Cheng, Zhouhui Lian, Sipin Nie, Longcun Jin, Gil Shamai, Yusuf Sahillioglu, Ladislav Kavan. An Evaluation of Canonical Forms for Non-rigid 3D Shape Retrieval. Graphical Models, 2018.  

Links and Downloads




This work was supported by EPSRC Research Grant EP/J02211X/1, National Natural Science Foundation of China Grant 61300135, Hong Kong Scholars Program Grant XJ2014058, Doctoral Fund of Ministry of Education of China Grant 20130172120001, Natural Science Foundation of Guangdong Province Grant S2013040016930, Open Research Fund of State Key Laboratory Grant I3I03, and TUBITAK under the project EEEAG-115E471.