A Computation-Aware Shape Loss Function for Point Cloud Completion

Authors: Shunran Zhang, Xiubo Zhang, Tsz Nam Chan, Shenghui Zhang, Leong Hou U

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experimental results confirm that our algorithm achieves the smallest gap with the real EMD within an acceptable time range and yields the best results in end-to-end training.
Researcher Affiliation Academia 1University of Macau 2Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences 3Shenzhen University
Pseudocode Yes The pseudocode for the specific algorithm is provided in the Appendix. Algorithm 1: Source Sorted Algorithm (SSA). Algorithm 2: Auction with Initial Price Algorithm.
Open Source Code Yes 1The code is available at https://github.com/coldbubbletea/ AAIP-Point-Cloud-Completion.
Open Datasets Yes In this study, the Shape Net CAD dataset (Chang et al. 2015) was used for spatial point cloud data.
Dataset Splits No The paper specifies training and testing data counts but does not explicitly mention a separate validation set or its split. 'Ultimately, we generated a total of 64000 pairs of point clouds for training and 9600 pairs of point clouds for testing.'
Hardware Specification No The paper does not provide specific hardware details (like GPU/CPU models, memory amounts, or detailed computer specifications) used for running its experiments.
Software Dependencies No The paper mentions software components and models (e.g., PCN, MSN) but does not provide specific version numbers for any key software dependencies or libraries.
Experiment Setup No The paper describes the adaptive iteration strategy with a formula but does not provide concrete hyperparameter values (e.g., learning rate, batch size, number of epochs, optimizer settings) or system-level training configurations for the deep learning models used in the experiments.