Recently, the paper entitled Learning to Incorporate Texture Salience Adaptive to Image Cartoonization, which was jointly written by Xiang Gao, Yuqi Zhang and Yingjie Tian, two students and a teacher of SEM, UCAS, was accepted by ICML 2022, a top level conference in the field of machine learning. The conference received a total of 5630 contributions, and 1117 were accepted, with an accepting rate of 19.8%.
The full name of ICML is International Conference on Machine Learning, an annual international top-level conference on machine learning sponsored by the International Machine Learning Society (IMLS). In the ranking of international academic conferences of the China Computer Federation (CCF), ICML is an A class conference in the field of artificial intelligence.
In this paper, a novel image animation style transfer model is proposed, and a parallel adversary learning of image level and region level is established. Region level branches constrain adversary learning to regions with significant cartoon image texture to better perceive and transmit cartoon texture features. This method can efficiently and deeply mine the texture features and surface features distribution of animation images, and achieve more significant and vivid style rendering effects on multiple datasets, especially high-resolution datasets. The research introduces a new idea for the study of cartoon style transfer.