VISION2UI
收藏资源简介:
VISION2UI是一个由华中科技大学等机构联合创建的真实世界数据集,专门用于从高保真设计图像自动生成UI代码。该数据集通过收集、清洗和过滤Common Crawl数据集中的数据构建而成,包含20,000个样本,每个样本包括设计图像和对应的UI代码。数据集创建过程中使用了神经评分器来保证数据质量,并通过人工标注的子集进行训练。VISION2UI数据集的应用领域主要集中在自动化UI代码生成,旨在简化开发流程,使初学者和设计师能够直接从设计图生成网页。
VISION2UI is a real-world dataset jointly created by Huazhong University of Science and Technology and other institutions, specifically designed for automatically generating UI code from high-fidelity design images. This dataset is constructed by collecting, cleaning and filtering data from the Common Crawl dataset, and contains 20,000 samples, each comprising a design image and its corresponding UI code. A neural scorer was employed during the dataset creation process to ensure data quality, and a manually annotated subset was used for training. The application fields of the VISION2UI dataset mainly focus on automated UI code generation, aiming to simplify the development workflow and enable beginners and designers to directly generate webpages from design drafts.

- 1VISION2UI: A Real-World Dataset with Layout for Code Generation from UI Designs华中科技大学 · 2024年



