Communications in Information and Systems

Volume 20 (2020)

Number 4

3D reconstruction using deep learning: a survey

Pages: 389 – 413

DOI: https://dx.doi.org/10.4310/CIS.2020.v20.n4.a1

Authors

Yiwei Jin (Department of Computer Science, Zhejiang University, Hangzhou, China)

Diqiong Jiang (Department of Computer Science, Zhejiang University, Hangzhou, China)

Ming Cai (Department of Computer Science, Zhejiang University, Hangzhou, China)

Abstract

Deep learning has remarkably improved the performance of many tasks in the computer vision community including 3D reconstruction. In this paper, we survey both classical and latest works of 3D reconstruction via deep learning. We divide all surveyed methods into three categories on the ground of the input modality: single RGB image based, multiple RGB images based and sketch based. Representations of output 3D shapes and specific goals of tasks are also taken into consideration in our classification. In addition, we overview datasets as well as evaluation metrics commonly used in current works. Finally, a discussion about potential directions of future research is provided.

The research is supported in part by National Natural Science Foundation of China (NSFC) (61972342) and the Science and Technology Department of Zhejiang Province (2018C01080).

Received 10 September 2020

Published 2 December 2020