MobiFace is a novel dataset for mobile face tracking and face grouping in the wild. It contains 80 unedited mobile live streaming recordings by smartphone users. More than 95K bounding box annotations are provided. Please note, we only accept requests from research or academic institutions. Any requests from free email addresses (hotmail, yahoo, gmail, 163, QQ etc) will be ignored.

If you find MobiFace useful, please consider citing:

@inproceedings{FT-RCNN,
    author = {Yiming Lin and Jie Shen and Shiyang Cheng and Maja Pantic},
    booktitle = {The IEEE International Conference on Automatic Face and Gesture Recognition (FG)},
    title = {FT-RCNN: Real-time Visual Face Tracking with Region-based Convolutional Neural Networks},
    year = {2020},
}

@inproceedings{mobiface_fg2019,
    author={Yiming Lin and Shiyang Cheng and Jie Shen and Maja Pantic},
    booktitle = {The IEEE International Conference on Automatic Face and Gesture Recognition (FG)},
    title={MobiFace: A Novel Dataset for Mobile Face Tracking in the Wild},
    year={2019},
    url={https://arxiv.org/abs/1805.09749v2}
}