pgu-nd / cg23-nervi

NeRVI: Compressive Neural Representation of Visualization Images for Communicating Volume Visualization Results

This repository contains the PyTorch implementation for paper "NeRVI: Compressive Neural Representation of Visualization Images for Communicating Volume Visualization Results".

Prerequisites

• Linux
• CUDA >= 10.0
• Python >= 3.7
• Numpy
• Pytorch >= 1.0

How to run the code

• First, change the directory path and iso values of the data in dataio.py.
• Second, set the parameter settings (e.g., batch size, lr etc.) in main.py.
• For training, set the 'train' as train and the 'approach' as CNN in the main.py.
• For testing, set the 'train' as inf in the main.py.

Citation

@article{gu2023nervi,
title={NeRVI: Compressive neural representation of visualization images for communicating volume visualization results},
author={Gu, Pengfei and Chen, Danny Z and Wang, Chaoli},
journal={Computers \& Graphics},
volume={116},
pages={216--227},
year={2023},
publisher={Elsevier}
}

Acknowledgements

This research was supported in part by the U.S. National Science Foundation through grants CNS-1629914, DUE-1833129, IIS-1955395, IIS-2101696, and OAC-2104158, and the U.S. Department of Energy through grant DE-SC0023145. The authors thank Ziang Tong for developing the visual interface.