Content-Conditioned Generation of Stylized Free hand Sketches

In recent years, the recognition of free-hand sketches has remained a populartask. However, in some special fields such as the military field, free-handsketches are difficult to sample on a large scale. Common data augmentation andimage generation techniques are difficult to produce images with variousfree-hand sketching styles. Therefore, the recognition and segmentation tasksin related fields are limited. In this paper, we propose a novel adversarialgenerative network that can accurately generate realistic free-hand sketcheswith various styles. We explore the performance of the model, including usingstyles randomly sampled from a prior normal distribution to generate imageswith various free-hand sketching styles, disentangling the painters' stylesfrom known free-hand sketches to generate images with specific styles, andgenerating images of unknown classes that are not in the training set. Wefurther demonstrate with qualitative and quantitative evaluations ouradvantages in visual quality, content accuracy, and style imitation onSketchIME.