Design And Implementation Of A New Animation Image Conversion System

With the development of animation and game industry in the new era,animation industry starts to be popular in people’s life.The current challenges facing anime character generation technology are also becoming increasingly serious,mainly in two aspects: 1)Unlike training face generation,the dataset for face generation has an innate pattern of five senses arrangement due to reality-based specificity,while the five senses in the secondary dataset are affected by the drawing method will be more different,which in turn affects the robustness of algorithm training.Previous methods basically do not use art standards to constrain the generated results.In addition,there is no dataset classified by painting style in the existing public dataset.2)The training algorithm requires a complex environment to be set up in advance and requires a certain level of programming knowledge,which makes it significantly more difficult for beginners to use.Therefore,for the above two problems from two perspectives of algorithm framework and system design,this paper proposes two solutions respectively.1: A new anime image conversion framework Fantastic Anime for cross-disciplinary research.its main contributions are as follows: 1)Research and organize a set of detection criteria that can be used in anime character generation techniques.Microsoft’s Media Pipe module is introduced to recognize anime character avatars and obtain the 3D coordinates of key points.The detection criteria are used to determine whether the generated anime avatars meet the requirements,which solves the problem of ambiguous rules when judging art standards.2)A new image conversion framework is designed.By integrating the generation and detection modules and adding the corresponding detection criteria in the detection module,this framework solves the problem of distorted and misaligned features when the trained model generates anime avatars.2: An anime avatar generation and detection system based on Fantastic Anime framework.Its main contributions are as follows.1)It can produce a large number of anime character avatars conforming to perspective within a short period of time.It solves the problem that a large number of anime avatars cannot be satisfied in a short period of time in game production and large online activities related to anime.2)It does not need to build an environment and can produce anime avatars by opening an applet.It solves the problems of complicated algorithm building environment and difficult model training for general animation enthusiasts.Through the mobile applet,this system allows users to quickly and easily complete anime avatar generation and detection.