HOPE: Shape Matching Via Aligning Different K-hop Neighbourhoods

Authors: Barakeel Fanseu Kamhoua, Huamin Qu

NeurIPS 2024 | 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, efficiency, and generalization ability of HOPE in the matching of nearly-isometric and non-isometric 3D shapes.
Researcher Affiliation Academia Barakeel Fanseu Kamhoua1, Huamin Qu1 , 1The Hong Kong University of Science and Technology
Pseudocode Yes Algorithm 1 : HOPE
Open Source Code Yes Justification: See Section 5 and attached code.
Open Datasets Yes We evaluate the performance of HOPE on two nearly isometric benchmark datasets TOSCA [7], and SCAPE [3], as well as on the non-isometric dataset SHREC 16 (TOPKIDS) [26], TOPKIDS
Dataset Splits No The paper discusses evaluation on "test pairs" and mentions using "all datasets" for parameter settings, but does not explicitly specify separate training, validation, and test dataset splits by name, or their sizes/percentages.
Hardware Specification Yes All experiments are conducted in Matlab 2023 on a Windows 11 system with 32GB RAM and Intel(R) i5 13500 CPU @ 2.50-4.8GHz.
Software Dependencies No The paper mentions 'Matlab 2023' but does not list other software dependencies with specific version numbers (e.g., libraries, frameworks).
Experiment Setup Yes On HOPE, on all datasets we set the LMD threshold ϵ, staring from ϵ = 100 and 10 equally spaced values to ϵ = 0.2 i.e., we use ϵ = linespace(100, 0.2, 10) and we set t = 60. When the last value of e is reached, it is maintained for the rest of the iterations. We equaly set kmax = 8 for all datasets.