Spec-Net and Spec-CGAN: Deep learning models for specularity removal from faces

作者:

Highlights:

• We introduce Spec-Face, the first public specularity data set for human faces.

• We introduce Spec-Net and Spec-CGAN, new deep models for face specularity removal.

• Spec-CGAN performs superior specularity removal for sample images from LFW dataset.

• Spec-Net performs superior specularity removal for Spec-Face and images from LIME.

• Classical methods perform poorly on low chromaticity faces.

摘要

•We introduce Spec-Face, the first public specularity data set for human faces.•We introduce Spec-Net and Spec-CGAN, new deep models for face specularity removal.•Spec-CGAN performs superior specularity removal for sample images from LFW dataset.•Spec-Net performs superior specularity removal for Spec-Face and images from LIME.•Classical methods perform poorly on low chromaticity faces.

论文关键词:Specularity,Dichromatic reflection model,Deep learning,Convolutional neural networks

论文评审过程:Received 1 August 2019, Revised 28 September 2019, Accepted 1 November 2019, Available online 7 November 2019, Version of Record 21 January 2020.

论文官网地址:https://doi.org/10.1016/j.imavis.2019.11.001