Unified GUI Grounding Dataset

该数据集包含了24,000个样本，这些样本来自包括移动设备、桌面和网页界面在内的多种平台，旨在提升图形用户界面（GUI）在不同分辨率下的定位能力。为了减少重复性，数据集采用了随机抽样方法，使得在较小的子集中也能实现与整体相当的性能表现。该数据集的任务是提高图形用户界面的定位准确性。This dataset contains 24,000 samples collected from multiple platforms including mobile devices, desktop environments, and web interfaces, with the goal of enhancing the positioning capability of graphical user interfaces (GUIs) across varying resolutions. To minimize redundancy, random sampling is employed during dataset construction, enabling smaller subsets to achieve performance comparable to that of the full dataset. The core task of this dataset is to improve the positioning accuracy of graphical user interfaces.

搜集汇总

背景与挑战

背景概述

该数据集是一个包含24,000个样本的图形用户界面（GUI）定位数据集，覆盖移动设备、桌面和网页界面等多种平台，旨在通过随机抽样减少重复性，提升GUI在不同分辨率下的定位准确性。