Datasets:
metadata
license: mit
task_categories:
- text-to-image
tags:
- art
size_categories:
- 10K<n<100K
Introduction
MangaZero dataset from paper DiffSensei: Bridging Multi-Modal LLMs and Diffusion Models for Customized Manga Generation
Please see GitHub repo to get the usage
Project page: https://jianzongwu.github.io/projects/diffsensei
The "type" key in character annotation
Type = 0: Characters with clean faces
Type = 1: Characters with vague faces
Type = 2: Exaggerated, artistic characters with appearances that don't conform to typical standards.