3D Car Shape Reconstruction from a Single Sketch Image

3D Car Shape Reconstruction from a Single Sketch Image

Abstract

Efficient car shape design is a challenging problem in both the automotive industry and the computer animation/games industry. In this paper, we present a system to reconstruct the 3D car shape from a single 2D sketch image. To learn the correlation between 2D sketches and 3D cars, we propose a Variational Autoencoder deep neural network that takes a 2D sketch and generates a set of multiview depth & mask images, which are more effective representation comparing to 3D mesh, and can be combined to form the 3D car shape. To ensure the volume and diversity of the training data, we propose a feature-preserving car mesh augmentation pipeline for data augmentation. Since deep learning has limited capacity to reconstruct fine-detail features, we propose a lazy learning approach that constructs a small subspace based on a few relevant car samples in the database. Due to the small size of such a subspace, fine details can be represented effectively with a small number of parameters. With a low-cost optimization process, a high-quality car with detailed features is created. Experimental results show that the system performs consistently to create highly realistic cars of substantially different shape and topology, with a very low computational cost.

Publication

Naoki Nozawa, Hubert P. H. Shum, Edmond S. L. Ho and Shigeo Morishima,
"3D Car Shape Reconstruction from a Single Sketch Image",
Proceedings of the 2019 International Conference on Motion, Interaction and Games (MIG) Posters

Best Poster Award

Downloads

YouTube

References

BibTeX

@inproceedings{nozawa193dcar,
 author={Nozawa, Naoki and Shum, Hubert P. H. and Ho, Edmond S. L. and Morishima, Shigeo},
 booktitle={Proceedings of the 2019 International Conference on Motion in Games},
 series={MIG '19},
 title={3D Car Shape Reconstruction from a Single Sketch Image},
 year={2019},
 pages={37:1--37:2},
 numpages={2},
 doi={10.1145/3359566.3364693},
 isbn={978-1-4503-6994-7},
 publisher={ACM},
 Address={New York, NY, USA},
 location={Newcastle upon Tyne, UK},
}

EndNote/RefMan

TY  - CONF
AU  - Nozawa, Naoki
AU  - Shum, Hubert P. H.
AU  - Ho, Edmond S. L.
AU  - Morishima, Shigeo
T2  - Proceedings of the 2019 International Conference on Motion in Games
TI  - 3D Car Shape Reconstruction from a Single Sketch Image
PY  - 2019
SP  - 37:1
EP  - 37:2
DO  - 10.1145/3359566.3364693
SN  - 978-1-4503-6994-7
PB  - ACM
ER  - 

Plain Text

Naoki Nozawa, Hubert P. H. Shum, Edmond S. L. Ho and Shigeo Morishima, "3D Car Shape Reconstruction from a Single Sketch Image," in MIG '19: Proceedings of the 2019 International Conference on Motion in Games, pp. 37:1-37:2, Newcastle upon Tyne, UK, ACM, 2019.

Similar Research

 

         

Last update: 02/07/2020