Face photo-sketch synthesis via full-scale identity supervision
作者：
Highlights：
• To the best of our knowledge, we first employ full-scale identity supervision to synthesis model, which provides a comprehensive constraint.
• Our CTIA divides the face synthesis into cross-domain translation and intra-domain adaptation, which is effective to generate realistic images.
• We first incorporate pseudo face images to augment the training set, and significantly improve the robustness for generative model.
摘要
•To the best of our knowledge, we first employ full-scale identity supervision to synthesis model, which provides a comprehensive constraint.•Our CTIA divides the face synthesis into cross-domain translation and intra-domain adaptation, which is effective to generate realistic images.•We first incorporate pseudo face images to augment the training set, and significantly improve the robustness for generative model.
论文关键词：Face photo-sketch synthesis,Identity supervision,Cross-domain translation,Intra-domain adaptation
论文评审过程：Received 11 November 2020, Revised 27 September 2021, Accepted 18 November 2021, Available online 22 November 2021, Version of Record 3 December 2021.
论文官网地址：https://doi.org/10.1016/j.patcog.2021.108446
