BRSU Skin/Face/Spoof Database

RGB- und SWIR-Portraits von Personen mit unterschiedlichen Hauttypen nach Fitzpatrick.

RGB and SWIR portraits of subjects with different skin types (classification by Fitzpatrick).

Background:

In the context of the research projects FeGeb and SPAI, skin samples from a large number of test subjects have been acquired. For this purpose, remission measurements of multiple positions on faces, arms and hands of the subjects have been recorded using a near / short wave infrared (NIR / SWIR) spectrometer. In addition, images of the subject’s faces have been captured from multiple angles using both a standard, but high-quality, RGB color camera and a multispectral SWIR camera system developed at the BRS-U. The primary goal of this measurement campaign was to proof the robustness of the method for contactless skin detection based on multispectral SWIR sensor and imaging systems proposed by the BRS-U. In the SWIR spectrum, human skin can easily be distinguished from other materials, regardless of age, gender or skin type. In addition, the acquired data is used to train classification algorithms to reliably distinguish “fakes”, such as latex masks, from genuine skin.

The resulting databases are partially available free of charge for academic and scientific purposes. This includes all 120 data sets of spectrometer measurements, as well as more than 50 data sets of RGB and SWIR images. The publication of these images has been approved by the pictured subjects. In addition, a database containing spectrometer measurements, RGB and SWIR images of various spoofing attacks such as masks and facial disguises is available.

Details of the database:

The BRSU multispectral skin/face database consists of two parts:

“Spectro”: Contains spectrometer measurements acquired from 120 subjects, measured at 16 positions on face, neck, arms and hands with a spectral range of 660nm to 1700nm using a irSys 1.7 spectrometer by TQ Group GmbH. These remission measurements have been performed using a special two-way lightguide and a standard Halogen lamp. For every data set, information about gender, age and skin type are provided.

“Face”: Contains color (RGB) and multispectral SWIR images consisting of the wavebands around 935nm, 1060nm, 1300nm and 1550nm, as well as the corresponding spectrometer measurements, of (currently) more than 50 subjects. The RGB images were taken with a Canon EOS 50D with 50mm lens. SWIR images were acquired with an Allied Vision Goldeye P032. The complete multispectral camera system is described in [1]. For every data set, information about gender, age and skin type are provided.

“Spoof”: Contains color (RGB) and multispectral SWIR images consisting of the wavebands around 935nm, 1060nm, 1300nm and 1550nm, as well as the corresponding spectrometer measurements, of various spoofing attacks, including (partial) masks and facial disguises. The RGB images were taken with a Canon EOS 50D with 50mm lens. SWIR images were acquired with an Allied Vision Goldeye P032. The complete multispectral camera system is described in [1].

Further details about the databases have been published by Steiner et al. (see [1] and [2]). While the Spectro database can be regarded as complete, the Face and Spoof databases might be extended in the future. Please check back here occasionally for information about updates.

Requesting the databases:

In order to request access to the databases, please complete the request form below (if you don’t see a form below, please follow this link). Provide your name, affiliation and a valid email address, ideally issued by your university, institute or organization. Please use the comments field to describe your field of research and the intended use of the database.

If your request is approved upon, you will receive an email containing information on how to download the databases.

The terms of use can be found here.

Current status:

“Spectro” is fully available.
“Face” was last updated on 2015-04-27.
“Spoof” was last updated on 2016-04-21.
There is no additional (incremental) update available yet.

Contact:

Prof. Dr.-Ing. Norbert Jung
Bonn-Rhein-Sieg University of Applied Sciences
Safety and Security Research Institute (ISF)
Grantham-Allee 2 – 8
53757 Sankt Augustin (Germany)
Tel: +49 (2241) 865 – 792
Fax: +49 (2241) 865 – 8792
email: isf-admin@h-brs.de

References:

[1]: Steiner, H.; Sporrer, S.; Kolb, A. & Jung, N.: “Design of an Active Multispectral SWIR Camera System for Skin Detection and Face Verification”, Journal of Sensors, vol. 2015, Article ID 456368.

[2]: Steiner, H.; Kolb, A. & Jung, N.: “Reliable Face Anti-Spoofing Using Multispectral SWIR Imaging”, Proc. Int. Conf. Biometrics (ICB), 2016 (accepted for publication).