Automatic Generation of Cinematic Animated Characters and Their Behavioral Characterization Using Graph Generation Networks

Abstract Animation is an art form that utilizes the phenomenon of visual transience to produce dynamic images through continuous filming, and in modern animation production, the design and creation of animation characters are often related to the popularity of animation works. This study responds to the development needs of the digital era and investigates the application value of image generation networks in the creation of movie-level animation characters. Through the analysis of the process of generating and creating animated characters, this paper constructs a method based on the graphic generation network to assist the creation and optimizes the image generation network model on the basis of deep learning. For the generated animation character action behavior problem, this paper also optimizes the visual semantic feature extraction based on behavioral feature extraction, so as to detect and study the abnormal behavior of the animation character and realize the detection of abnormal behavior of the generated animation character, so as to provide a guarantee for the subsequent animation production. Through experiments, it is found that the node degree distribution, clustering coefficient distribution, and average track count distribution of this paper’s model on the COKK dataset are 0.178, 0.185, and 0.076, respectively, and all of them achieve the best results. The average recognition accuracy of abnormal behaviors in animated characters is as high as 96.76%. The experimental results verify that the methods for animated character generation and abnormal behavior detection in this paper have certain effectiveness and feasibility and provide a reference for modernizing the animation industry.