Overview of VR Behavior Dataset

The Head and Gaze VR Behavior Dataset is a large-scale dataset containing both the head motion and eye gaze information.

• It contains nearly 20 million pieces of head and gaze tracking data, represented as coordinate and quaternion along with corresponding images.
• The data sequences are captured among 100 users with a high sampling frequency 120Hz.
• It provides 27 videos, which have a rich diversity based on three technique metrics: camera motion, video quality, and the dispersion of region of interest (ROI).

Our dataset can support many tasks in 360° videos, for example:

• 360° Video Streaming
• User Identification
• Psycho Analysis
• Video Recommdentation

Other Dataset

Here are disparities between our dataset and exist several datasets that provide head movement tracks of users.
Our dataset outperforms on the length of dataset, number of participants and frequency.
Our dataset have both head and gaze data and more abundent diversity of contents.

Table 1: Existing datasets.

Publications

The following paper describes VR Behavior dataset in depth: from the data processing to detailed statistics about the data. If you use the VR Behavior dataset in your work, please cite this article.

Yili Jin*, Junhua Liu*, Fangxin Wang, Shuguang Cui. Where Are You Looking?: A Large-Scale Dataset of Head and Gaze Behavior for 360-Degree Videos and a Pilot Study ACM Multimedia, 2022. [arXiv] [BibTeX]

@inproceedings{vrdataset,
  author = {Yili Jin and Junhua Liu and Fangxin Wang and Shuguang Cui},
  title = {Where Are You Looking?: A Large-Scale Dataset of Head and Gaze Behavior for 360-Degree Videos and a Pilot Study},
  booktitle = {{MM} '22: Proceedings of the 30th ACM International Conference on Multimedia, Lisboa, Portugal, October 10 - 14, 2022},
  publisher = {{ACM}},
  year = {2022},
  doi = {10.1145/3503161.3548200}
}

Videos

We collect videos from YouTube, and the following table shows the URL and content of the videos, where the full URL is the concatenation of "www.youtu.be/" and the string listed in the table.

Table 2: Information of videos