Education, Science, Technology, Innovation and Life
Open Access
Sign In

Research of 3D Virtual Characters Reconstructions Based on NeRF

Download as PDF

DOI: 10.23977/jeis.2023.080606 | Downloads: 17 | Views: 281


Lingyi Song 1


1 Chongqing No.8 Secondary School, No.8 Gongyuanbei Road, Yubei District, Chongqing, China

Corresponding Author

Lingyi Song


With the development of the Internet, Metauniverse, and graphics processing technique, video games and immersive virtual social contacting service are largely propelled. Therefore, the quality of 3D virtual character models is getting more and more important—higher fidelities of screens have largely promoted the demand for finer 3D character models. However, the cost and time efficiency of traditional modeling relied on human artist can hardly support such a great demand, and will potentially slow down the development of video game and metauniverse industry. In an effort to improve this situation, this paper conducted research about 3D virtual characters reconstructions through NeRF and illustrated the principals and functions of NeRF. Using two different datasets (Doll Photos Dataset and Real-Human Photos Datasets), this paper evaluated the NeRF model and provided researching advice for future research about dataset building and possible directions.


Neuron Radiance Field; 3D-Reconstruction; Human Reconstruction


Lingyi Song, Research of 3D Virtual Characters Reconstructions Based on NeRF. Journal of Electronics and Information Science (2023) Vol. 8: 41-54. DOI:


[1] China Audio-video and Digital Publishing Association (GPC). Chinese Gaming Industry Report 2022. February 2, 2023.
[2] China Audio-video and Digital Publishing Association (GPC). Gametech – A New Technology Cluster in The Process of Digital-real Integration. July 21, 2022.
[3] Diana Werner, Ayoub Al-Hamadi, Philipp Werner. Truncated Signed Distance Function: Experiments on Voxel Size. University of Magdeburg. 2014.
[4] Zhenbao Liu, Hongliang Qin, Shuhui Bu, Meng Yan, Jinxin Huang, Xiaojun Tang, Junwei Han. 3D Real Human Reconstruction Via Multiple Low-cost Depth Cameras. Northwestern Polytechnical University. 2017.
[5] Matthew Loper, Naureen Mahmood, Javier Romero, Gerard Pons-Moll, Michael J. Black. SMPL: A Skinned Multi-person Linear Model. 2015.
[6] Nelson Max. Optical Models for Direct Volume Rendering. University of California, Davis, and Lawrence Livermore National Laboratory. 1995.
[7] Ben Mildenhall, Pratul P. Srinivasan, Matthew Tancik, Jonathan T. Barron, Ravi Ramamoorthi, Ren Ng. NeRF: Representing Scenes as Neuron Radiance Fields for View Synthesis. UC Berkeley, Google Research, and UC San Diego. 2020.
[8] Yao Yao, Zixin Luo, Shiwei Li, Jingyang Zhang, Yufan Ren, Lei Zhou, Tian Fang, Long Quan. BlendedMVS: A Large-scale Dataset for Generalized Multi-view Stereo Networks. The Hongkong University of Science and Technology, Everest Innovation Technology, and Zhejiang University. 2019.
[9] Diederik P. Kingma, Jimmy Lei Ba. Adam: A Method for Stochastic Optimization. 2017.

Downloads: 6319
Visits: 248849

Sponsors, Associates, and Links

All published work is licensed under a Creative Commons Attribution 4.0 International License.

Copyright © 2016 - 2031 Clausius Scientific Press Inc. All Rights Reserved.