Abstract
We present the first real-world dataset and quantitative evaluation of visual attention of mobile device users in-situ, i.e. while using their devices during everyday routine. Understanding user attention is a core research challenge in mobile HCI but previous approaches relied on usage logs or self-reports that are only proxies and consequently do neither reflect attention completely nor accurately. Our evaluations are based on Everyday Mobile Visual Attention (EMVA) – a new 32-participant dataset containing around 472 hours of video snippets recorded over more than two weeks in real life using the front-facing camera as well as associated usage logs, interaction events, and sensor data. Using an eye contact detection method, we are first to quantify the highly dynamic nature of everyday visual attention across users, mobile applications, and usage contexts. We discuss key insights from our analyses that highlight the potential and inform the design of future mobile attentive user interfaces.
Download
To access the dataset, please download, fill in, and sign the End User License Agreement (EULA). The EULA has to be signed by a representative from your organisation (students are not allowed). Please use your official email address when requesting access to the dataset. Once the form has been filled in, please send your request to emva.dataset@gmail.com.
The Dataset shall only be used as mandated by the End User License Agreement. If you use this dataset in any of your work, please cite the following paper:
- Mihai Bâce, Sander Staal, and Andreas Bulling. 2020. Quantification of Users’ Visual Attention During Everyday Mobile Device Interactions. In Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems (CHI ’20). Association for Computing Machinery, New York, NY, USA, 1–14. DOI:https://doi.org/10.1145/3313831.3376449
@inproceedings{bace20_chi,
title = {Quantification of Users' Visual Attention During Everyday Mobile Device Interactions},
author = {Bâce, Mihai and Staal, Sander and Bulling, Andreas},
year = {2020},
booktitle = {Proc. ACM SIGCHI Conference on Human Factors in Computing Systems (CHI)},
doi = {10.1145/3313831.3376449}
}
EMVA Dataset
The figure below shows the structure of our dataset.
Each participant's data is stored in a separate directory named using the participant's ID. Inside each folder, there will be several Recordings (also directories) and additional metadata.
- Recording: a recording began when the study participants explicitly started the data collection application and it finished also when participants stopped or paused the data collection. This was a privacy requirement imposed by the Ethics Committee from ETH Zürich.
- Session: a session is part of a recording. It started when a user unlocked their device until the user locked their device again or the device went into standby. Due to file size constraints, the duration of a sesion was limited to around 15 mins. Longer interactions from the user are stored as several consecutive sessions.
- Each session contains several files that store each data source. The video.mp4 file stores the video recording corresponding to that session. Not all the mobile devices had the same capabilities. In this case, the corresponding file will be empty (e.g. most phones did not have a humidity sensor).
We are continously working to improve this website and the description of the dataset.
Code Samples
To help others get started with the dataset, we provide a few code samples on GitHub: Code Repository
Contact
If you have any questions, feel free to reach out: Mihai Bâce (mihai.bace@inf.ethz.ch)
Acknowledgements
Andreas Bulling was supported by the European Research Council (ERC; grant agreement 801708).