National Key Research and Development Project (2016YFB1001200); National Natural Science Foundation of China (61872346)
Free-hand sketches have always been one of the important tools for human communication. As it can express some complex human thoughts quickly in a succinct form, the study of free-hand sketches has always been one of the research hotspots in the field of computer vision. Currently, the research on free-hand sketches mainly focuses on the recognition, retrieval and completion. As researchers focus on the fine-grained operation of free-hand sketches, research on free-hand sketch segmentation has also received more and more attention. In recent years, with the development of deep learning and computer vision technology, a large number of free-hand sketch segmentation methods based on deep learning have been proposed. Moreover, the accuracy and efficiency of free-hand sketch segmentation have also been significantly increased. However, free-hand sketch segmentation is still a very challenging topic because of the abstraction, sparsity and diversity of free-hand sketches. At present, there are few Chinese reviews on hand-drawn sketch segmentation. This paper organizes, categorizes, analyzes and summarizes the free-hand sketch segmentation algorithm based on deep learning to solve the above deficiency. Firstly, Show three basic sketch representation methods and commonly used sketch segmentation datasets. According to the sketch segmentation algorithm prediction results, introduce sketch semantic segmentation, sketch perceptual grouping and sketch parsing respectively. Moreover, Collect and arrange the evaluation results of sketch segmentation on the primary data sets. Finally, summarize the application of sketch segmentation and discuss the possible future development direction.