High Resolution Range based Face Database (HRRFaceD)
A face database is presented composed by a set of high resolution range images acquired by the latest generation of range / depth cameras: the Microsoft Kinect 2 (second generation).
The database is composed by the faces of 18 people, acquired from different poses: frontal, lateral, etc. The faces of some of the people have been acquired with and without glasses.
The database is structured in different folders as:
/01 (person with identifier 01)
-->/01/test (images for testing the person with identifier 01)
-->/01/test/01_009.png, ..., 03_096.png, ... (images starting with the same identifier as the root folder, 01 in this case, are the true/positive samples, and the others are the false/negative samples)
-->/01/train (images for training the person with identifier 01)
-->/01/train/01_034.png, ..., 05_g_168.png, ... (images starting with the same identifier as the root folder, 01 in this case, are the true/positive samples, and the others are the false/negative samples)
/04_g (person with glasses who, in addition, is the same as the person in /04)
Every root folder (01, 02, ...) contains the range images of one person. The folder name is the identifier of the person (for example 01, 02, etc.). There are some folders whose name end in '_g' (for example 04_g). This suffix indicates that the person wears glasses, but he/she is really the same person that the other existing folder with the same numeric identifier but without the '_g' suffix. For example, the folders '04' and '04_g' correspond to the same person without and with glasses. Inside every root folder, there are two folders: test (for testing purposes) and train (for training purposes).
Inside every subfolder, there are a set of range face images that can be true/positive samples or false/negative samples. True/positive samples have a name that starts by a numeric identifier that is the same as the corresponding root folder, and false/negative samples have a name that starts by a different identifier (for instance, an image in '/01/test/01_009.png' is a true/positive sample, but an image in '/01/test/03_096.png' is a false/negative sample).