ICBA Face Verification Contest on
the BANCA dataset
Organised by Kieron Messer, Josef Kittler and Rachel Gartshore
University of Surrey, Guildford, Surrey. GU2 7XH, UK.
In order to promote evaluation of pattern recognition algorithms using publicly available datasets and standard performance assessment methodology, a competition for the best face verification algorithms took place in conjunction with the ICBA 2004 conference.
The tests were carried out using the English images acquired from the BANCA database. A brief description of the database was presented at AVBPA2003 (Fourth International Conference on Audio and Video-based Biometric Person Authentication) and is available on-line from here. Data from the database are available at the cost of distribution from here.
The BANCA Protocol
The experiments were carried out according to the BANCA protocol
using the Matched Controlled (MC) configuration. There were two separate parts to the competition.
Part I: Pre-registered Images
Images were supplied which have already been localised (by hand) and geometrically normalised. The resulting resolution of the images is 55x51 pixels and 8-bit grey-scale and contains just face pixels. These images were then be used to train and test the authentication system. To enter this part of the contest the false acceptance and false rejection rates have to be submitted to us on configuration MC of the BANCA protocol at three different operating points on a ROC curve.
Part II: Automatic Registration
Full video resolution colour images, each containing a face were supplied. All participants had to use an automatic method of localisation for the at least the test phase of the protocol. You may use manual localisation in the training phase. Eye co-ordinates for the training images were supplied. Again, the error rates (i.e. FA and FR) were required for entrance into this part of the competition at the three different operating points.
- The results to the competition were published in the conference proceedings. An on-line version of the paper can be download here.
- Also, a standard set of face recognition software from the internet was used to provide a baseline performance measure. Due to space limitations the results of these were not published in the proceedings but can be found here.
For more information, contact Kieron Messer