3D Reconstruction Algorithms and Comparison

PubDate: July 2022

Teams: GANTU Technology CO Ltd;Nanjing University of Posts and Telecommunications

Writers: Zhimeng Chi; Xinru Wang

PDF: 3D Reconstruction Algorithms and Comparison

Abstract

In the previous several years, the field of 3D reconstruction and generation has been growth rapidly and produced multiple approaches for different sub-tasks such as representing 3D structure, convolutional operations or generating latent vector for feature extraction. With the aim of surveying approaches in different conceptions, we have classified and studied six representative articles published in 2018 to 2021. This article focuses on discuss how the different representations or convolution process effect on the performance and output shapes. For each method, it covers their main motivations and attributions, the conceptions or architecture of the learning model, their experiment datasets and approach, and the result of their works. In the next part, we go through the analysis at the present stage of related research and discuss the potential future works inspired by potential optimization. This work is an effort to comprehensive research of current approach of 3D reconstruction task, and it can provide reference guide for those who are interested in such field.