The Hong Kong Polytechnic University Hyperspectral Face Database (PolyU-HSFD)
The Biometric Research Centre (UGC/CRC) at The Hong Kong Polytechnic University has developed a Hyperspectral Face database to advance research and to provide researchers working in the area of face recognition with an opportunity to compare the effectiveness of face recognition algorithms. The indoor hyperspectral face acquisition system was built which mainly consists of a CRI’s VariSpec LCTF and a Halogen Light (Illustrated in Fig. 1), and includes a hyperspectral dataset of 300 hyperspectral image cubes from 25 volunteers with age range from 21 to 33 (8 female and 17 male). For each individual, several sessions were collected with an average time space of 5 month. The minimal interval is 3 months and the maximum is 10 months. Each session consists of three hyperspectral cubes-- frontal, right and left views with neutral-expression. The spectral range is from 400nm to 720nm with a step length of 10nm, producing 33 bands in all. Fig. 2 shows an example of the hyperspectral face. Since the database was constructed over a long period of time, significant appearance variations of the subjects, e.g. changes of hair style and skin condition, are presented in the data. In data collection, positions of the camera, light and subject are fixed, which allows us to concentrate on the spectral characteristics for face recognition without masking from environmental changes.
Fig. 1 The established hyperspectral face imaging system
Fig. 2 Illustration of a set of 33 hyperspectral face bands
Each Mat file is 3-D datacube with size: 220 (height) *180 (width) *33 (no. of bands), cropped from the original images, which contains the radiance data from the original source .tif image.
Three types of images were acquired: front, right, and left. The Mat-files are stored separately as:
Front-Mat file: in the file of SampleImages_F, totally 151 datacubes.
Right-Mat file: in the file of SampleImages_R, totally 125 datacubes.
Left-Mat file: in the file of SampleImages_L, totally 124 datacubes.
(B) Source Data
Source data are documented according to the acquisition date. Totally four data sets (I to IV) were acquired at different time (10/17/2007, 11/21/2007-11/23/2007, 12/12/2007, and 3/21/2008). Some subjects may have several sessions at each time. Each session (e.g. Sample 17 from the 3/21/2008 data set) contains 3 sub-files: f-front image, r-right image and l-left image.
For a detailed and completed data description, you are advised to read the DATA_TEXT and DATA_TABLE. Also, we manually located the eye coordinates from the front images (151 subjects) for registration and cropping. Here is the format:
F_1_1.jpg 61 93 118 89
F_1_1.jpg: the name of the front face image
61: the X-coordinate of the right eye (from the view of the sample face, R_x)
93: the Y-coordinate of the right eye (from the view
of the sample face, R_y)
118: the X-coordinate of the left eye (from the view of the sample face, L_x)
89: the Y-coordinate of the left eye (from the view of the sample face, L_y)
1. Wei Di, Lei Zhang, David Zhang, and Quan
Hyperspectral Face Recognition in Visible Spectrum
with Feature Band Selection, IEEE Trans. on System, Man
and Cybernetics, Part A, vol. 40, issue 6, pp. 1354
-1361, Nov. 2010 .
The Announcement of the Copyright
All rights of the PolyU-HSFD are reserved. The database is only available for research and noncommercial purposes. Commercial distribution or any act related to commercial use of this database is strictly prohibited. A clear acknowledgement should be made for any public work based on the PolyU-HSFD. A citation to "PolyU-HSFD Database, http://www.comp.polyu.edu.hk/~biometrics/hyper_face.htm" and our related works must be added in the references.
Here we would like to thank Chen Chao and Yang Hong fang’s kind help for the data acquisition.
Downloading Steps (Recommend to open by IE):
This PolyU Hyperspectral Face
Database ("hsface.zip") is made publicly available and can be obtained from this website. It is
totally free for academic, noncommercial purposes. Any person or organization
that wishes to use the database must agree to the terms of the agreement and
fill in the agreement forms. The
successful applicants will receive their login accounts and passwords to
download the database.
Biometric Research Centre, The