ARQGAN: An evaluation of generative adversarial network approaches for automatic virtual inpainting restoration of Greek temples

作者：

Highlights：

• A dataset formed by images of 3D models of Greek temples has been created.

• Two Deep Learning approaches for automatic image restoration of ruined temples.

• Both approaches have been evaluated with mathematical metrics with good results.

• Restored images have been evaluated by people in the field of architecture.

摘要

•A dataset formed by images of 3D models of Greek temples has been created.•Two Deep Learning approaches for automatic image restoration of ruined temples.•Both approaches have been evaluated with mathematical metrics with good results.•Restored images have been evaluated by people in the field of architecture.

论文关键词：Deep Learning,Generative Adversarial Networks,Image inpainting,Segmented training,Virtual restoration,Greek temples

论文评审过程：Received 4 October 2020, Revised 8 March 2021, Accepted 19 April 2021, Available online 22 April 2021, Version of Record 10 May 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2021.115092