Enhancing Artistic Style Transfer: Integrating CycleGAN, Diffusion Models, and Neural Painting for Monet-Inspired Image Generation

This project explores various approaches to style transfer, specifically focusing on transforming natural images into the iconic style of Claude Monets paintings. We explored three methods: CycleGAN-based, neural painting-based, and diffusion- based. CycleGAN enables style transfer between domains without paired training data, neural painting simulates the physical painting process, and diffusion models leverage a denoising process for high-quality results. To evaluate the effectiveness of these approaches, we conduct a survey assessing the aesthetic appeal, naturalness, and adherence to Monets style of the generated images. Our analysis provides insights into the strengths and limitations of each method and identifies areas for future improvement in reproducing Monets iconic style.