A Novel Vector Sketch Generation Method Based on Adversarial Autoencoder

Aiming at the problems of existing vector sketch generation methods,such as generated sketch scribble and single coding sketch information,this paper proposed a novel vector sketch generation method based on adversarial autoencoder(called Sketch-AAE).Sketch-AAE takes advantage of the adversarial mechanism to merge the spatial information of a raster sketch into the vector sketch generation,which makes the generated sketch having better category shape information.The proposed method utilizes not only the temporal information among the strokes in vector sketch,but also the spatial information of the object in raster sketch.The extensive sketch generation experiments were conducted on the QuickDraw dataset,and the Ske-score was used as the quantitative measurement.The experimental results show that the proposed method can alleviate the scribble effect of the generated sketches and achieves better visual impression and higher category discrimination.