Enhancing single-view 3D mesh reconstruction with the aid of implicit surface learning

作者：

Highlights：

• A hybrid approach toward single-view 3D object reconstruction.

• A two-branch network is proposed to reconstruct both implicit and explicit shapes.

• The two representations are forced to be consistent through a surface matching loss.

• The final shape is smoother and has lower surface-to-surface distance error.

摘要

Highlights•A hybrid approach toward single-view 3D object reconstruction.•A two-branch network is proposed to reconstruct both implicit and explicit shapes.•The two representations are forced to be consistent through a surface matching loss.•The final shape is smoother and has lower surface-to-surface distance error.

论文关键词：Geometric deep learning,Single-view reconstruction,Mesh reconstruction,Implicit surface reconstruction,Hybrid approach

论文评审过程：Received 31 October 2021, Revised 8 January 2022, Accepted 11 January 2022, Available online 19 January 2022, Version of Record 31 January 2022.

论文官网地址：https://doi.org/10.1016/j.imavis.2022.104377