Synthetic Anime Manga Pairs

This dataset contains synthetic image pairs for anime-to-manga translation tasks. Each pair includes a colored anime-style image and a corresponding manga-styled version, generated using the framework proposed in the paper “A Two-Stage Unsupervised GAN Approach for Anime-to-Manga Translation.”

Input images were selected from the Danbooru dataset using specific tags to ensure diversity and to avoid copyright concerns. The manga images were created through a two-step process: image segmentation followed by the application of screen patterns, simulating traditional manga styles while preserving structure and visual coherence.

The dataset is intended for research purposes such as training and evaluating anime-to-manga models, screen tone generation methods, and visual style transfer techniques. For further details, please refer to the associated paper and its authors.