Face sketch synthesis and reputation have wide range of packages in law enforcement. Despite the amazing progresses had been made in faces cartoon and reputation, maximum current researches regard them as separate responsibilities. On this paper, we propose a semantic neural version approach so that you can address face caricature synthesis and recognition concurrently. We anticipate that faces to be studied are in a frontal pose, with regular lighting and neutral expression, and have no occlusions. To synthesize caricature/image photos, the face vicinity is divided into overlapping patches for gaining knowledge of. The size of the patches decides the scale of local face systems to be found out.

翻译：人脸素描合成与识别在执法领域具有广泛应用。尽管人脸漫画与识别已取得显著进展，但现有研究大多将二者视为独立任务。本文提出一种语义神经网络方法，可同时处理人脸素描合成与识别任务。我们假设所研究的人脸图像为正脸姿态、具有均匀光照和中性表情，且无遮挡。为合成素描/照片图像，将人脸区域划分为重叠的局部块进行学习，这些局部块的尺寸决定了待学习的局部人脸结构尺度。