ElonMusk-v2/GroundCUA

GroundCUA是一个大型且多样化的数据集，包含真实的UI截图与结构化标注，用于构建多模态计算机使用代理。它覆盖了87个软件平台，涵盖生产力工具、浏览器、创意工具、通信应用、开发环境和系统工具，支持Windows、macOS、Linux和跨平台应用。数据集包括JSON标注文件和PNG截图，通过SHA-256哈希配对，标注包含UI元素的边界框坐标、文本内容和粗略语义类别（如按钮、菜单等）。该数据集专为GUI grounding、UI感知和视觉语言动作模型的研究设计，适用于训练计算机代理理解UI布局、构建GUI grounding模块、预训练屏幕解析器以及基准测试OCR和布局分析等任务。

GroundCUA is a large and diverse dataset of real UI screenshots paired with structured annotations for building multimodal computer use agents. It covers 87 software platforms across productivity tools, browsers, creative tools, communication apps, development environments, and system utilities, spanning Windows, macOS, Linux, and cross-platform apps. The dataset includes JSON annotation files and PNG screenshots paired via SHA-256 hashes, with annotations containing bounding boxes, text, and coarse semantic categories for UI elements. It is designed for research on GUI grounding, UI perception, and vision-language-action models, supporting use cases such as training agents to perceive UI layouts, building GUI grounding modules, pretraining screen parsers, and benchmarking OCR and layout analysis.