DatAndroid

Motivation

DatAndroid is a benchmark dataset dedicated to the development and assessment of Visual GUI Testing (VGT) techniques in the Android domain. VGT techniques are based on image recognition techniques that are able to identify the same widgets rendered across multiple devices. With this goal in mind, DatAndroid was mined to illustrate the development of two features matching-based approaches for widget detection in the field of visual GUI testing with the purpose to evaluate the portability of widget over different apps and devices.

DATA SOURCE

The dataset  contains 95 apps downloaded from Uptodown emulated on 14 different devices from  "Writing and Notes" category uptodown store, containing mostly utilities to manage and organize text-based content and item lists. Datas available show screenshotsFor each device and app, data  and associated metadata (View Hierarchies) are included. The dataset is organized as follows:

• Device emulated
  • App Folder  
    • Screenshot.png 
    • metadata.xml

DEVICE EMULATED

• Nexus 5 (1080 × 1920)
• Pixel 3 (1080 × 2160)
• Pixel 2 (1080 × 1920)
• Nexus 5X (1080 × 1920)
• Nexus 6P (1440 × 2560)
• Nexus 6 (1440 × 2560)
• Nexus S (480 × 800)
• Pixel 3a (1080 × 2220)
• Pixel 3a XL (1080 × 2160)
• Pixel (1080 × 1920)
• Nexus 4 (768 × 1280)
• Pixel 3 XL 1440 × 2960)
• Pixel 2 XL (1440 × 2880)
• Pixel XL (1440 × 2560)

CITATION:

@article{ardito2021feature,
  title={Feature Matching-based Approaches to Improve the Robustness of Android Visual GUI Testing},
  author={Ardito, Luca and Bottino, Andrea and Coppola, Riccardo and Lamberti, Fabrizio and Manigrasso, Francesco and Morra, Lia and Torchiano, Marco},
  journal={ACM Transactions on Software Engineering and Methodology (TOSEM)},
  volume={31},
  number={2},
  pages={1--32},
  year={2021},
  publisher={ACM New York, NY}
}