Face Sketch Synthesis: A Neural Style Approach

Face sketch synthesis plays a significant role in the domain of face recognition and retrieval applications. Great interest of its applications in law enforcement and entertainment have continued to be noted in the research community. Over the years, great advances have been noted in face sketch synthesis as most recent methods are gradually moving towards deep representation based approaches. In this paper, we propose a style transfer approach for synthesis of sketches by leveraging recent algorithms applied to artistic style transfer. In addition, the approach shows the application style transfer produces visually appealing synthetic sketches results as well as facial reconstructions. Lastly, we show poorly reconstructed images can be improved with a denoising autoencoder to create better synthetic images. The effectiveness of the approach is presented through experiments on the CUHK face sketch database.