Boota-S2A: A Multi-Modal Synthetic Dataset for Sketch-Based 3D Anime Character Reconstruction

Boota-S2A is a multi-modal synthetic dataset designed for sketch-based 3D anime character reconstruction research. The dataset contains 120 synthetic anime-style 3D character models in GLB format. For each character model, five rendered views are provided: Back, Front, Left, Right, and ThreeQuarter. Each rendered view is paired with two sketch-style representations: a colored sketch and a colorless sketch. The dataset includes the following components: - 120 anime-style 3D character models in GLB format;- multi-view rendered images generated from the 3D models;- colored sketch-style images generated from the rendered images;- colorless sketch-style images generated from the rendered images;- a metadata.csv file that provides file-level correspondence between each 3D model, rendered view, colored sketch, and colorless sketch. The original anime-style characters were created by the dataset authors using VRoid Studio. The models were converted to GLB format and rendered using Blender/Python scripts. Colored and colorless sketches were generated from the rendered images using Python image processing methods. No original VRM files are distributed in this dataset. The released 3D models are provided only in GLB format. Boota-S2A v1.0 is intended for non-commercial academic and research purposes only. Commercial use of the 3D models, rendered images, colored sketches, and colorless sketches is not permitted without written permission from the dataset authors. If you use this dataset, please cite the related paper and the Zenodo dataset record.