A Video Dataset for Multiview Face Localization and Tracking

Junya Sato and Takuya Akashi  
Smart Computer Vision Laboratory (Akashi Labo)
Iwate University, Japan


Videos and ground truth 1 (800 MB)     Videos and ground truth 2 (923 MB)         

Method overview

Method overview

 

 

About this dataset
 This dataset provides multiview face detection, localization, and tracking researchers a collection of videos in which each subject drastically moves the face. The above figures show all the first frames of each video sequence. The dataset consists of two types of datasets. In the first dataset ((a) to (c)), 10 subjects with three different backgrounds are captured by HD Pro Webcam C910 (Logicool) with 30 fps. In the second dataset ((d) to (i)), five subjects, four of them are the same as the first dataset, with six different backgrounds are captured. The first face direction is arbitrary and the subjects freely and drastically move the face by the end. The speed of the movement is not restricted, however it is not fast to avoid blur. Each video consists of about 200 images of bmp.

 

Method overview

 

 

Ground truth of a rotated minimum bounding box
 This dataset provides the ground truth of the rotated mininum bounding box as shown above figures. The ground truth includes face parts, such as palpebral fissures, nares, and a mouth. The ground truth is provided as csv files in which image names, x and y coordinates of upper left, upper right, lower right, and lower left are described.

 

Method overview Method overview Method overview

 

 

Localization and tracking results by the proposed method
 Above figures show the examples of the localization and tracking result by the proposed method.

 

 

 

Citation
 If you use this dataset, please cite the following paper:
[1] @article{junya2016,
title={High-speed Multiview Face Localization and Tracking with a Minimum Bounding Box Using Genetic Algorithm},
author={Junya Sato and Takuya Akashi},
journal={IEEJ Transactions on Electrical and Electronic Engineering},
volume={??},
number={??},
year={2016},
publisher={IEEJ}
}

 

 

 

Copyright

THIS DATA SET IS PROVIDED "AS IS" AND WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.

The images provided above may have certain copyright issues. We take no guarantees or responsibilities, whatsoever, arising out of any copyright issue. Please use it at your own risk.