A pipeline that generates user-preferred, photo-realistic avatars using Stable Diffusion and Bayesian Optimization.
Unlike traditional avatar generators, where users manually adjust each attribute, our pipeline allows the program to determine the most suitable avatar based on user ratings. Our goal is to simplify the avatar creation process by focusing more on user preferences and less on parameter adjustments. Our pipeline utilizes stable diffusion to synthesize high-resolution, realistic avatars. Furthermore, Bayesian optimization plays a crucial role in fine-tuning the generated avatars according to specific user preferences, incorporating feedback loops to iteratively impove towards a user-preferred avatar.
The project is open-sourced and available at GitHub. Feel free to contact us.
