AMEX

该数据集名为AMEX，包含了超过10.4万个高分辨率的应用程序截图，这些截图来自110款热门移动应用。数据集在多个层面上进行了注释，包括图形用户界面（GUI）交互元素的定位、GUI屏幕及元素功能描述，以及复杂的自然语言指令。此外，该数据集还包括三个级别的注释，旨在通过提供复杂的自然语言指令，增强通用型移动GUI控制代理的能力。该数据集规模宏大，截图数量超过10.4万张，其任务是用以训练和评估移动GUI控制代理。

This dataset, named AMEX, contains over 104,000 high-resolution mobile application screenshots collected from 110 popular mobile apps. It is annotated at multiple levels, covering the localization of graphical user interface (GUI) interactive elements, functional descriptions of GUI screens and their corresponding elements, as well as complex natural language instructions. Additionally, the dataset includes three levels of annotations, which aim to enhance the capabilities of general-purpose mobile GUI control agents by providing complex natural language instructions. With a large-scale volume of over 104,000 screenshots, this dataset is specifically designed for training and evaluating mobile GUI control agents.