Saturday October 21, 2017

You may also be interested in:
The XM2VTS score-level Fusion Benchmark Database
The Biosecure DS2 score+quality Database
Return to the main page

Blind Subject Face database (BSFD)

Using your face to unlock a mobile device is not only an appealing security solution, but also a desirable or entertaining feature, such as taking selfies (self image). It is convenient, fast, and does not require much effort, but only if you have no vision problems. For visually impaired users, taking selfies could potentially be a challenge. In order to study the usability and ensure the inclusion of mobile-based identity authentication technology, we have collected the Blind-subject Face DataBase (BSFD).

The BSDB was gathered in the St Nicholas Home centre, which is a centre for visually impaired people (Penang, Malaysia). The scenario was a quiet room where each participant was only accompanied by the operator and the ambient conditions were normal (e.g. lightning, temperature and humidity). In all the experiments the user was sat at a table and was requested to hold the camera by himself and try to focus his face in order to obtain a face image as centred as possible. The process was scheduled in two sessions and each one consisted of four different experiments: E1 (no feedback), E2 (audio feedback), E3 (instructions) and E4 (feedback and instructions).

The camera used for capturing the face images is an Advent Slim 300K web cam connected to a PC. The images taken have a 640x480 resolution in grayscale. Each experiment consists of the user taking a selfie video and within 45s) in order to obtain a good image of his face. Each video is divided into 400 snapshots and there are 8 videos per user (4 experiments in each session) and 40 users: in total 128000 timestamped snapshots are acquired. However, not all the snapshots contain valid facial images. There are visually impaired users, consisting of 29 men and 11 woman, with the following age distribution: 45% are under 25 years old, 42% are between 25 and 50, and 13% are over 50. 15 of them have low vision, 15 can distinguish light and 10 are completely blind.

A data collection session.

Example of images

some examples of images

If you do use this database, please cite the following publication:

Poh, Norman; Blanco-Gonzalo, Ramon; Wong, Rita; Sanchez-Reillo, Raul: 'Blind subjects face database', IET Biometrics, 2016, 5, (1), p. 20-27, DOI: 10.1049/iet-bmt.2015.0016, IET Digital Library [link]


Download the dataset here.


Please come back soon!

Norman Poh
CVSSP, SEPS, University of Surrey Guildford, Surrey, GU2 7XH, United Kingdom.