Exact Shape Correspondence via 2D graph convolution

Authors: Barakeel Fanseu Kamhoua, Lin Zhang, Yongqiang Chen, Han Yang, MA Kaili, Bo Han, Bo Li, James Cheng

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We report experimental results that validate the effectiveness and efficiency of 2D-GEM on exact matching of non-isometric and nearly-isometric 3D shapes.
Researcher Affiliation Academia 1The Chinese University of Hong Kong, 2The Hong Kong University of Science and Technology, 3Hong Kong Baptist University
Pseudocode Yes Algorithm 1 : 2D-GEM and Algorithm 2 : 2-Hop
Open Source Code Yes Code at: https://github.com/Barakeel Fanseu/2D-GEM
Open Datasets Yes We conduct experiments to evaluate the performance of our method on two nearly isometric benchmark datasets TOSCA [10], and SCAPE [5], as well as a benchmark non-isometric dataset SHREC 16 (TOPKIDS) [42].
Dataset Splits No The paper states that the method needs no training ('Given that our method needs no training, is unsupervised...', Section 7). Therefore, no traditional validation dataset split is used or specified.
Hardware Specification Yes Time comparison between 2D-GEM and GRAMPA is conducted in Matlab on a Windows 10 system with 16GB RAM and Intel(R) i5 11400 CPU @ 2.60-4.4GHz.
Software Dependencies No The paper mentions 'Matlab' but does not provide specific version numbers for Matlab or any other software libraries or dependencies used in the experiments.
Experiment Setup Yes For 2D-GEM, we initialize the correspondence by the GMWM on SHOT [70], and use e = 10... For nearly-isometric shapes... we use 500 eigen-vectors of the LBO... set the maximum iteration number to iter = 10. We also set the LMD thresholds, we use [0.5280, 0.4560, 0.3840, 0.3120, 0.2400, 0.1680, 0.1680, 0.1680, 0.1680, 0.1680]. For non-isometric shapes (TOPKIDS), we set the thresholds to 100 for all iterations, use iter = 60, and 10 LBO eigen-vectors.