Idiap Research Institute
Centre du Parc
Rue Marconi 19
PO Box 592
CH - 1920 Martigny
T +41 27 721 77 11
F +41 27 721 77 12
Welcome to Idiap Research Institute's biometric resources page. Below, you will find useful protocols, annotations, etc. that are provided to help encourage reproducible research.
There are even more resources available via http://www.idiap.ch/resource/faceverif.
SCface - Surveillance Cameras Face Database (link)
- Face Authentication Protocol based on the DayTime Tests (download) (bibtex)
This protocol for the SCface database was developed by Mitchell McLaren and Roy Wallace to allow for testing of face verification/authentication algorithms in challenging real-world surveillance/forensics conditions, with substantial lighting, pose and image quality variation. The protocol is based based on the DayTime tests scenario, and utilises images from all 130 subjects in the database. It includes world (for training), development (for tuning) and test sets each with a disjoint set of subjects. More details are available via scface.org and in these papers: IJCB paper; IEEE Trans. IFS article.
MOBIO Database (link)
The MOBIO database contains videos of 150 participants captured in challenging real-world conditions on a mobile phone camera over one and a half years. The protocols, similar to those described for the 2010 ICPR MOBIO Face and Speaker Verification Competition are distributed with the database.
- MOBIO Still-Image Database and Face Verification Protocol (bibtex)
To create the MOBIO still-image database one image was extracted from each of the videos and was manually annotated with eye locations. Using manual face localisation allows for the evaluation of face authentication accuracy separately from the choice of face detection algorithm. These images are used to evaluate face verification/authentication systems, following the standard MOBIO protocol. Thanks are due to Niklas Johansson for completing the laborious task of manually annotating the images that now form the MOBIO still-image database. You can download the images and protocol via the MOBIO database webpage. More information can be found in the following papers: IJCB paper; IEEE Trans. IFS article.
CMU Multi-PIE Face Database (link)
- Multi-PIE Face Verification Protocol - Unmatched Illumination (download) (bibtex)
The CMU Multi-PIE Face Database consists of images from 337 subjects captured in up to four different sessions over a six month period. Images exhibit variation through differences in pose, illumination and facial expressions. As a face verification/authentication protocol was not publicly available for Multi-PIE, we developed our own protocol with independent world, development and evaluation sets, designed to study the effect of illumination variation on face verification accuracy. More information can be found in the following papers: IEEE Trans. PAMI article; IET Biometric article.
Extended training data sets
- BMFM training data set protocol (download) (bibtex)
For experimental purposes we created a training data set referred to as BMFM, composed of images from the training sets of the BANCA, MOBIO, FRGC and MultiPIE databases. The result is a training set of 32440 images of 510 subjects that contains a wide range of subject and session variability. When using BMFM in our experiments, a universal background model (UBM) was trained from a randomly-selected subset of 1275 images of 425 subjects (3 images each), while the full BMFM training set was used for training other parameters.
Labeled Faces in the Wild Database (link)
You can download the annotations here.
This dataset contains i-vectors of natural and synthetic speech utterances for spoofing and anti-spoofing research purposes
download: (download TIFS2015.zip)
This dataset contains score files resulting from experiments described in the Springer book chapter Face Recognition in Challenging Environments: An Experimental and Reproducible Research Survey
download: (download FRICE.zip)
CVPR Biometric Workshop 2016
This dataset contains score files resulting from experiments described in the paper Heterogeneous Face Recognition using Inter-Session Variability Modelling
download: (download CVPRW_2016.tar.gz)
The following tarball contains score files obtained with PAD experiments reported in the paper Cross-database evaluation of audio-based spoofing detection systems
download: (download interspeech_2016.tar.gz)
The following tarballs contain feature-files for image-quality measures (IQM) for the Replay-Mobile and Replay-Attack databases, as described in the paper: The REPLAY-MOBILE Face Presentation-Attack Database.
download files for Replay-Mobile database: (download biosig2016_replaymobile_iqm_features.tar.gz)
download files for Replay-Attack database: (download biosig2016_replayattack_iqm_features.tar.gz)
The following tarballs contain score files obtained with CNN-based PAD systems reported in the paper: On the Use of Convolutional Neural Networks for Speech Presentation Attack Detection.
download scores (200 MB): (download isba2018-pad-dnn.tar.gz)