MUD

MUD数据集由莫纳什大学创建，旨在为现代风格的用户界面（UI）建模提供一个大规模且经过噪声过滤的数据集。该数据集包含18,132个独特的UI屏幕，每个屏幕都附有高质量的视图层次结构，这些数据是从3.3k个Android应用中自动挖掘出来的。MUD数据集通过使用LLMs模拟人类探索方式，自动从应用中提取UI数据，并结合成熟的噪声过滤技术和人工验证确保数据质量。该数据集主要用于UI元素检测和UI检索等任务，旨在为高质量、现代UI的研究提供基础。The MUD dataset was created by Monash University, aiming to provide a large-scale, noise-filtered dataset for modern-style user interface (UI) modeling. This dataset contains 18,132 unique UI screens, each paired with high-quality view hierarchies, and the data was automatically mined from 3.3k Android applications. The MUD dataset automatically extracts UI data from applications by using LLMs to simulate human exploration patterns, and combines mature noise filtering techniques and manual verification to ensure data quality. This dataset is mainly used for tasks such as UI element detection and UI retrieval, aiming to provide a foundation for research on high-quality, modern user interfaces.

搜集汇总

背景与挑战

背景概述

MUD是由莫纳什大学创建的大规模现代UI数据集，包含18,132个独特UI屏幕及其视图层次结构，源自3.3k个Android应用的自动挖掘。它通过LLMs模拟人类探索结合噪声过滤和人工验证确保高质量，主要用于UI元素检测和检索任务，为现代UI研究提供基础。