Multi-instance Sketch To Image Synthesis Based On Labeled Sketch

The painting design process involves a lot of creative work.This process usually begins with a sketch on paper,where designers and engineers share their ideas and try to express the real scenes of the artwork.Real-world images usually contain multiple objects,as a result,generating an image from a multi-instance sketch is an attractive research topic which has more practical significance.In addition,with the development of data-driven intelligent era,acquiring labeled simulation sample data can effectively compensate for the lack of real data,so we propose a multi-class sketch image generation scheme based on labeling.On the one hand,it can assist the creative work of art practitioners,on the other hand,it solves the difficult problem of labeling data manually.However,existing generative networks usually produce a similar texture on different instances for those methods focus on learning the distribution of the whole image so that ignore the distribution of different categories instances.To address this problem,we propose a progressive instance texture reserved generative approach to generate more convincible images by decoupling the generation of the instances and the whole image.Specifically,we create an instance generator to synthesize the primitive color distribution and the detailed texture for each instance.Then,an image generator is designed to combine all of these instances to synthesize an image retaining texture and color.Besides,to generate more significant details,such as eyes,ears etc.,we propose a novel technique called Discriminative Sketch Augmentation,which can provide structural constraint by obtaining the sketch of the discriminative region.Extensive experiments demonstrate that our model not only generates convincing images but also achieves higher Inception Score and lower Fr0)?(8?0) inception distance on the MSCOCO dataset.This means that our scheme achieves good results in image quality and diversity.We collect a high quality dataset that provides abundant multi-instance sketch training data for the multi-instance image generation.The dataset is also suitable for other tasks of sketch images,such as sketch classification,sketch recognition,etc.We collect a high quality dataset that provides abundant multi-instance sketch training data for the multi-instance image generation.The dataset is also suitable for other tasks of sketch images,such as sketch classification,sketch recognition,etc.In summary,our scheme has practical value in the sketch to image task.