Abstract
This paper proposes a Generative Adversarial Network (GAN)-based approach for generating anime-style cartoon avatars, addressing the current limitations in cartoon avatar datasets, including their scarcity and copyright restrictions. Due to the limited availability of high-quality, diverse cartoon datasets, particularly for anime-style faces, obtaining large-scale labeled data for training deep learning models remains a significant challenge. To overcome these issues, we train a GAN model on a large collection of anime-style images to generate unique, high-quality cartoon faces. The generator learns to capture the distinctive features of anime art, such as exaggerated expressions, vibrant colors, and stylized facial proportions, while the discriminator ensures the realism and diversity of the generated avatars. Our experimental results demonstrate that the proposed method not only produces visually appealing avatars faithful to the anime aesthetic but also offers a scalable solution to the datasets limitations and copyright concerns. This approach opens up new possibilities for applications in digital art, gaming, and social media, where custom cartoon avatars are increasingly in demand.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call