NeuralIndicator: Implicit Surface Reconstruction from Neural Indicator Priors

Authors: Shi-Sheng Huang, Guo Chen, Chen Li Heng, Hua Huang

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

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive evaluations on synthetic and real-scan datasets show that our approach consistently outperforms previous approaches, especially when point clouds are incomplete and/or noisy with complex topology structure.
Researcher Affiliation Academia 1School of Artificial Intelligence, Beijing Normal University, Beijing, China.
Pseudocode No The paper describes methods and processes in textual form and through mathematical equations, but it does not include structured pseudocode or algorithm blocks.
Open Source Code No The paper includes footnotes referencing supplementary materials for results and computational cost analysis, but it does not provide an explicit statement or link for open-source code availability for the methodology described.
Open Datasets Yes We use five public point cloud datasets, including both synthetic datasets (ABC (Koch et al., 2019), FAMOUS (Erler et al., 2020), Reconbench (Berger et al., 2013), Thingi10K (Zhou & Jacobson, 2016)) and real-scan dataset (DPoint (Wu et al., 2015) dataset)...
Dataset Splits No The paper mentions using specific test subsets from public datasets for evaluation but does not explicitly provide details on training or validation splits with percentages, counts, or a clear splitting methodology for those subsets.
Hardware Specification Yes It takes about one hour for our system to finish 10, 000 training epochs on a desktop computer with an NVIDIA Ge Force RTX 3060 12G GPU.
Software Dependencies No The paper does not explicitly provide specific version numbers for software dependencies used in the experiments.
Experiment Setup Yes Specifically, we set voxel resolution as 32, 64, 128 and 256 voxels in the unit box, and perform the surface reconstruction with such four system variants (called SIFG-32, SIFG-64, SIFG-128 and SIFG-256 respectively) on the Test dataset.