takara-ai/fudeno-instruct-4M

Fudeno-Instruct-4M是一个包含超过400万个多模态数据条目的综合数据集，旨在训练模型理解和生成SVG图形。数据集由25%的随机生成的SVG和75%的高质量人工curated的SVG组成，涵盖了从简单形状到复杂设计的各种SVG复杂性。适用于文本到SVG生成、多模态矢量图形理解、视觉元素代码生成和研究结构化图形表示等研究和发展。Fudeno-Instruct-4M is a comprehensive dataset containing over 4 million multimodal data entries designed to train models for understanding and generating SVG graphics. The dataset consists of 25% randomly generated SVGs and 75% high-quality curated SVGs, covering a range of SVG complexities from simple shapes to intricate designs. It is suitable for research and development in text-to-SVG generation, multimodal understanding of vector graphics, code generation for visual elements, and understanding structured graphic representations.