Reproduce “AnimeDiffusion: Anime Face Line Drawing Colorization via Diffusion Models” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yu Cao, Xiangqiao Meng, P.Y. Mok, Xueting Liu, Tong-Yee Lee, Ping Li
AnimeDiffusion introduces a diffusion-based model for automated anime face line drawing colorization, leveraging reference images to guide the process. The approach demonstrates superior performance over existing GAN-based methods in quantitative metrics such as LPIPS, FID, and SSIM, while also showing improved color accuracy and semantic correspondence in user studies.
Key insights
- Diffusion models offer a stable and effective alternative to GANs for specialized image-to-image translation tasks like anime colorization.
- A hybrid training approach that combines general denoising with specific image reconstruction objectives improves model convergence and final output quality.
- Applying geometric transformations like Thin-Plate Splines (TPS) to reference images during training significantly enhances the model's ability to maintain semantic correspondence across varying poses and expressions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.