A modified convolutional neural network for face sketch synthesis
作者：
Highlights：
• A modified convolutional neural network for face sketch synthesis is proposed.
• This method is free of complex optimization problems and pre/post-processing compared with existing methods.
• We examine the trade-off between the model and performance and in the limited case the proposed method degrades into a conventional CNN architecture.
• It improves the synthesis efficiency by one or two orders of magnitude and achieves highly competitive numerical results.
• The proposed method is more robust to inhospitable conditions and can be easily extended to RGB channels.
摘要
•A modified convolutional neural network for face sketch synthesis is proposed.•This method is free of complex optimization problems and pre/post-processing compared with existing methods.•We examine the trade-off between the model and performance and in the limited case the proposed method degrades into a conventional CNN architecture.•It improves the synthesis efficiency by one or two orders of magnitude and achieves highly competitive numerical results.•The proposed method is more robust to inhospitable conditions and can be easily extended to RGB channels.
论文关键词：Face sketch synthesis,Convolutional neural networks,Sparse coding
论文评审过程：Received 9 November 2016, Revised 28 August 2017, Accepted 18 October 2017, Available online 28 October 2017, Version of Record 6 November 2017.
论文官网地址：https://doi.org/10.1016/j.patcog.2017.10.025
