Abstract
In this paper, we propose a method for reconstructing the 3D model from a single 2D image. The current cutting-edge methods for 3D reconstruction use the GAN (Generative Adversarial Network) to generate the model. However, the methods require multiple 2D images to reconstruct the 3D model, because all the information of a real object cannot be obtained from only one side, especially the back of the object is invisible. Since rebuilding a 3D model from a single view is an important issue in practical applications, the system requires the ability to obtain information about the surrounding environment of the object more quickly without the need for the object to move around.
Therefore, we propose a method for reconstructing 3D models from an image by learning the relationship between 3D model and 2D image. Mainly this method consists of three parts. The first part is the view layer, observing real-world objects and capturing 2D images. The layer searches the related 2D image of the 3D model that exists in the 3D model library. The second part is the corresponding layer. The 2D image corresponding to the 3D model is taken out, and contrast with real-world 2D images of objects. The 2D cross-section of the 3D model is found as the most similar one to the 2D image of the real-world object. The third part is generative layer that based on the model library to find the corresponding 3D model, reconstructing a 3D model that corresponds to the real object by using the GAN.
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsReferences
Yang, B., Wen, H., Wang, S., Clark, R., Markham, A., Trigoni, N.: 3D object reconstruction from a single depth view with adversarial learning. In: ICCV Workshops (2017)
Li, Y., Dai, A., Guibas, L., Nießner, M.: Database-assisted object retrieval for real time 3D reconstruction. In: Computer Graphics Forum (2015)
Fan, H., Su, H., Guibas, L.J.: A point set reconstruction network for 3D object reconstruction from a single image. http://www.arxiv.org/
Gadelha, M., Maji, S., Wang, R.: 3D shape induction from 2D views of multiple objects. In: CVPR (2017)
Choy, C.B., Xu, D., Gwak, J., Chen, K., Savarese, S., 3D-R2N2: a unified approach for single and multi-view 3D object reconstruction. http://www.arxiv.org/
Dou, P., Shah, S.K., Kakadiaris, I.A.: End-to-end 3D face reconstruction with deep neural networks. http://www.arxiv.org/
Chen, Z., Tong, Y.: Face super-resolution through Wasserstein GANs. http://www.arxiv.org/
Tatarchenko, M., Dosovitskiy, A., Brox, T.: Machine learning multi-view 3D models from single images with a convolutional network. http://www.arxiv.org/
Wu, J., Zhang, C., Xue, T., Freeman, B., Tenenbaum, J.: Learning a probabilistic latent space of object shapes via 3D generative-adversarial modeling. In: NIPS (2016)
Smith, E., Meger, D.: Improved adversarial systems for 3D object reconstruction and reconstruction. In: CORL (2017)
Wu, J., Wang, Y., Xue, T., Sun, X., Freeman, B., Tenenbaum, J.: MarrNet: 3D shape reconstruction via 2.5D sketches. In: NIPS (2017)
Qi, C. R., Su, H., Nießner, M., Dai, A., Yan, M., Guibas, L.J.: Volumetric and multi-view CNNs for object classification on 3D data. http://www.arxiv.org/
Author information
Authors and Affiliations
Corresponding authors
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2018 Springer Nature Switzerland AG
About this paper
Cite this paper
Si, L., Yasumura, Y. (2018). Reconstruction of a 3D Model from Single 2D Image by GAN. In: Kaenampornpan, M., Malaka, R., Nguyen, D., Schwind, N. (eds) Multi-disciplinary Trends in Artificial Intelligence. MIWAI 2018. Lecture Notes in Computer Science(), vol 11248. Springer, Cham. https://doi.org/10.1007/978-3-030-03014-8_20
Download citation
DOI: https://doi.org/10.1007/978-3-030-03014-8_20
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-03013-1
Online ISBN: 978-3-030-03014-8
eBook Packages: Computer ScienceComputer Science (R0)