VISION2UI
收藏资源简介:
VISION2UI数据集,由华中科技大学、北京大学、重庆大学等机构联合创建,旨在提升多模态大型语言模型(MLLMs)在用户界面(UI)设计图像生成代码方面的的能力。该数据集包含20000个样本,其中16,000个用于训练模型,2000个用于验证,2000个用于测试,每个样本包含设计图像、UI代码及其布局信息。数据来源于Common Crawl开源数据集,经过收集、清洗、筛选等一系列操作,确保了数据的高质量和真实性,并通过训练神经网络评分器对数据进行进一步筛选,保留了更高质量的实例。该数据集的构建,不仅为自动化UI代码生成的研究提供了宝贵的资源,也为初学者和设计师直接从设计图生成网页提供了可能,具有重要的应用价值和市场潜力。
The VISION2UI dataset, jointly created by institutions including Huazhong University of Science and Technology, Peking University, Chongqing University and other organizations, aims to enhance the capabilities of multimodal large language models (MLLMs) in generating code for user interface (UI) design images. This dataset contains 20,000 samples in total, with 16,000 allocated for model training, 2,000 for validation, and 2,000 for testing. Each sample includes design images, UI codes and their corresponding layout information. The data is sourced from the Common Crawl open-source dataset, and has undergone a series of processes such as collection, cleaning and screening to ensure high data quality and authenticity. Furthermore, a neural network-based scorer is employed to further filter the dataset, retaining higher-quality instances. The construction of this dataset not only provides a valuable resource for research on automated UI code generation, but also enables beginners and designers to directly generate web pages from design images, holding significant application value and market potential.




