Publication

Abstract : Cartoons are a ubiquitous art form in day-today living, and auto-generating cartoons from photos are highly desirable. But cartoons demand smooth color shading, clear edges, and moderately basic surfaces, which pose critical challenges for texture-descriptor-based capacities as cartoonification and its styles demand intriguing characteristics with high levels of disentanglement and reflection. GAN-based learning techniques are effectively instituted for converting real-world photographs or scenic images into cartoon-style images. But not all the styles and cortoonification methods are suitable, as a few applications demand versatile generative models while others require pure cartoon-style transformations. The proposed work primarily focuses on the implementation feasibility of various GAN approaches, such as CartoonGan and StyleGan2. Further, these models are analyzed and evaluated for their utility potential and variances in different image and cartoon styles, so that suitable models are recommended for the needs of Industry 5.0 and Society 5.0 applications. © 2024 IEEE