File(s) not publicly available
VGG-360 Face
Version 3 2020-03-04, 13:56
Version 2 2019-05-15, 14:45
Version 1 2019-05-15, 14:31
dataset
posted on 2020-03-04, 13:56 authored by Marcos CirneMarcos Cirne, Fernanda AndaloFernanda Andalo--------------------------------
VGG-360 Face Dataset
--------------------------------
Description:
This dataset uses images from the VGG Face Dataset of Oxford University (http://www.robots.ox.ac.uk/~vgg/data/vgg_face) to simulate, for each face image, 10 different projections on a 360º camera. This is done by using camera angle information from MOT-360 Face Dataset (which can be downloaded at https://figshare.com/s/5f8a69937014e4720625) and the pixel coordinates of the eyes of each face image. As these coordinates are not provided by the original authors, the Dlib Face Detector (http://dlib.net/) was used in this process. However, the detector could not be able to detect all faces from the VGG Face Dataset. Therefore, approximately 750K from this dataset images were used to simulate the projections, yielding a total of 7.5M images in this dataset.
Citation:
When referencing the dataset, please use the following BibTex citation:
@inproceedings{Cirne2019,
author = "M. Cirne and F. Andalo and R. Dias and T. Resek and G. Bertocco and R. Torres and A. Rocha",
title = "Deep Face Verification for Spherical Images",
booktitle = "{IEEE} International Conference on Image Processing",
year = "2019"
}
