Face recognition has witnessed a great deal of
awareness due to its numerous applications in many fields like
computer vision, security, pattern recognition and computer
graphics, but still is a challenging and active research area. In
this paper, we have presented a comprehensive survey of face
databases for constrained and unconstrained Environments.
Face databases are utilized for the face detection and
identification algorithm testing and they have been formulated
to evaluate the effectiveness of face recognition algorithms.
The paper is focused mostly on novel databases that are freely
availablefor the research purposes. Most of the popular face
databases are briefly introduced and compared.
Keywords- face recognition, face database, expression, occlusion.
I. INTRODUCTION
Over the past few years research in recognition of faces has
moved from 2D to 3D. The demand for 3D face data has
evolved in the need of 3D databases. In this work, we first
give an introduction of publicly accessible 2D and 3D face
databases for constrained and unconstrained Environments.
The numerous existing databases requires a quantitative
analysis of these databases so as to compare more accurately
the performances of the different algorithms present in
literature. The development of algorithms robust to
illumination, pose, facial expression, age, occlusion changes
des demands databases of adequate size that involves
carefully controlled changes of these features. Also,
conventional databases are needed to relatively analyze these
methods.
Presently, many existing databases are utilized for
recognition of faces that differs in lighting conditions, size,
pose, expressions, the quantity of pictured people and
occlusions. The prior facial databases largely consists of
frontal images. Nowadays, the facial databases were seen to
take into account the changes in pose, illumination, imaging
directions, ethnicity, gender and facial expressions. Few of
the most recent databases take into account the changes in
picture sizes, compression, occlusions and are collected from
different sources like social media and web.
In this paper, we have presented a comprehensive survey
of face databases for constrained and unconstrained
environments. Section II describes theoverview of various
face databases. It focuses mostly on novel databases that are
freely available for research purposes. Section III describes
some of the recent face databases. Section IV compares the
various popular face databases. Finally, Section V describes
the conclusion of the paper.
II. FACE DATABASES
Over the past few decades, numerous face databases have
been designed to analyze the effectiveness of face recognition
algorithms. The brief introduction of selected databases is as
follows. In most cases the link to database download is
provided.
A. The AR database
It is a database which has genuine occlusions and are open to
the public. It has above 4,000 color pictures of 126 people
faces (70 men and 56 women). These pictures undergo
through wide changes in facial expressions, lighting
conditions and occlusions like sunglasses and scarves. They
were captured under strictly controlled environments. There
were no restrictions imposed on wear of people (clothes,
glasses, etc.), makeup, hair style, etc. For every subject, 26
images in total were captured in two sessions (two weeks
apart) 1.
The limitations of the AR database are that it only contains
two types of occlusions, i.e., sunglasses and scarf, and the
location of the occlusion is either on the upper face or lower
face. It is available on the link
http://rvl1.ecn.purdue.edu/~aleix/aleix face DB.html.
Fig.1 Sample images of two sessions from AR database 2.
B. The Extended Yale B database
It consists of 2,414 frontal face pictures of 38 persons in 64
different lighting conditions. For each person in a typical
pose, an picture with surrounding (background) illumination
was also recorded. The pictures are grouped into four
subsets according to the lighting angle in concern to the axis
of camera. The Subset 1 and Subset 2 cover the angular
range 0? to 25?, the Subset 3 extends from 25? to 50?, the
Subset 4 covers 50? to 77?, and the Subset 5 covers angles
which are larger than 78?. In order to simulate various levels
of contiguous occlusions, the most used scheme is to restore
a randomly placed square patch from every test picture with
a baboon picture that has analogous texture with the human
face. The position of the occlusion is randomly chosen. The
sizes of the synthetic occlusions vary in the range of 10% to
50% of the original image 2.
Fig.2 Occlusion variations images from the Extended Yale B database 2.
C. The FRGC database
The well-known 3D expression databases are the “Face
Recognition Grand Challenge” (FRGC) databases. It had a
high footprint on the advancement of face recognition
techniques. So it is also considered as the reference databases
for validating the 3D face recognition techniques. The Face
Recognition Grand Challenge (FRGC) database has 8,014
pictures from 466 people in difference sessions. For each
subject in each session, there are four controlled stationary
pictures, two unconstrained stationary pictures, and a single
3D picture. The still images contain variations such as
lighting and expression changes, time-lapse, etc.
The unconstrained pictures were captured in changing
illuminations. Every set of unconstrained images contains
two expressions, smiling and neutral. To simulate the
randomly located occlusions, one can restore a randomly
located square patch from every picture by a black block. The
location of the occlusion is randomly selected. The size of the
black block varies in the range of 10% to 50% of the original
image 2. It is available on the link
https://www.idiap.ch/software/bob.
D. The LFW database
It is a database of face photographs designed for analyzing
the issue of unconstrained recognition of faces which
contains 13,233 face
Fig.3 Sample images from the LFW database: first and second row: six
matched pairs from six subjects, third and forth row: six non-matched
pairs from twelve subjects 2.
pictures of 5,749 people gathered from the web. These
pictures are taken in uncontrollable conditions that has major
changes in pose, expression, lighting, time-lapse and different
types of occlusions. These faces are only identified using the
Viola-Jones face detector. Every face has been labelled by the
name of the person pictured. All the 1,680 people pictured
have 2 or more variant images in the database. The aim of face
verification under the LFW database’s protocol is to verify if a
pair of face pictures belongs to the same subject or not. The
pictures are present as 250 x 250 pixel JPEG pictures. In this
database, most of pictures are in color, though some are
grayscale only 4.
E. CAS-PEAL Database
It consist of images from 66-1040 people (595 men, 445
women) in 7 different groups: pose, expression, accessory,
Fig. 4 Pose variation in the CAS-PEAL database 2.
time, lighting, background, and distance. In this database, for
the pose subset, nine cameras located in a semicircle around
the people were utilized. Pictures were captured continuously
within a small time span (2 seconds) 2. It is available on
the link http://www.jdl.ac.cn/peal/index.html.
F. FERET
It was a collective work of Dr. Wechsler and Dr. Phillips. The
pictures were gathered in a semi-controlled condition. In order
to keep a degree of uniformity in the database, the same
physical arrangement was utilized in all the photography
session. As the equipment had to be assembled again for every
session, there were few minor changes in pictures collected on
dissimilar dates.
It has 1564 sets of pictures for a total of 14,126 pictures that
consists of 1199 people and 365 duplicate sets of pictures. A
duplicate set is another set of pictures of an individual
already present in the database and was generally captured on
a dissimilar day. It is available on the link
http://www.nist.gov/humanid/feret/. The color FERET
dataset can be downloaded from the link
http://www.nist.gov/humanid/colorferet/.
Fig.5 pose variations images from FERET database 6.
G. Korean Face Database (KFDB)
It consists of facial images of plenty of Korean people
gathered under constrained environments. In this, pictures
with changing pose, lighting, and facial expressions were
captured. The people were photographed in the mid of an
octagonal frame and the cameras were located between 450
off frontal in two orientations at 150 increments.
Fig. 6 Pose variation images from Korean face database 2.
H. Yale Face Database B
It was the database gathered for the efficient testing of face
recognition techniques under considerable changes in
lighting and pose. The people were pictured inside a geodesic
dome by 64 computer-controlled xenon strobes. The pictures
of 10 people were captured under 64 lighting conditions in 9
different poses. It is available on the link
http://cvc.yale.edu/projects/yalefacesB/yalefacesB.html.
Fig. 7 Illumination variations from the Yale Face Database B 6.
I. Yale Face Database
It is a database which consists of 11 pictures of 15 people in a
different environments having with and without glasses,
variations in facial expression and lighting variation 2. This
database can be downloaded from the link
http://cvc.yale.edu/projects/yalefaces/yalefaces.html.
J. CMU Pose, Illumination, and Expression (PIE) Database
This database efficiently samples a plenty of pose and
illuminations and different facial expressions. It has created an
influence on algorithm advancement for recognition of faces
across pose. It consists of 41,368 pictures captured from 68
people. The RGB color pictures are 640×480 in size 3. This
database can be downloaded from the link
http://www.ri.cmu.edu/projects/project 418.html.
Fig. 8 Illumination variation images from PIE face database 3.
K. SCface Database
This database consists of stationary pictures of human faces.
The pictures were captured in unconstrained indoor conditions
utilizing 5 video surveillance cameras of different features. It
has 4160 stable pictures of 130 people. The pictures from
variant quality cameras creates the practical-world
environments which helps in robust testing of algorithms for
recognition of faces. It is freely available to research
community 3.
L. Georgia Tech Face Database
It consists of pictures of 50 individuals which are present in
JPEG format. In this database, most of the pictures were
captured in two dissimilar sessions to consider the changes in
lighting conditions, facial expression, and appearance. Also,
the faces were taken at dissimilar scales and directions. All
pictures are is manually marked to find the location of the face
in the picture.
M. Japanese Female Facial Expression (JAFFE) Database
It is a database which consists of 213 pictures of seven facial
expressions (6 normal + 1 neutral) in various poses by ten
Japanese female models. In this, all pictures has been
evaluated on 6 emotion features by 60 Japanese people 2. It
can be downloaded from the link
http://www.mis.atr.co.jp/˜mlyons/jaffe.html.
Fig. 9 Expression variation images from JAFFE database 2.
N. Indian Face Database
It has eleven different pictures of 40 different people. All the
pictures are stored in JPEG format. The size of every picture
is 640 by 480 pixels, which are having 256 grey levels per
pixel. The pictures are arranged in 2 main categories – males
and females. The different directions of the face included are:
looking front, left, right, up, down, up towards left, up
towards right and the different emotions present are: smile,
neutral, laughter, sad/disgust 3. It is available on the link
http://www.cs.umass.edu/~vidit/facedatabase.
Fig. 10 Pose variation images from Indian Face database 3.
O. FEI Face Database
It is a database of Brazilian faces which consists of a set of
face pictures captured at the Artificial Intelligence
Laboratory of FEI in Brazil. It has 14 pictures of all the 200
people, a 2800 images in total.
P. The Bosphorus database
It is a novel 3D face database that has a large set of
expressions, efficient variations of poses and various types of
occlusions. It is very useful for the advancement and analysis
of techniques on recognition of faces under dreadful
environments, facial expression evaluation and synthesis.
Q. FaceScrub Database
The database was gathered from the pictures present on the
Web. It has an automatic procedure to justify that the picture
associates to the right person. It has the pictures of 530
persons, a total of 107,818 available in the database. The
pictures are provided with the name and gender annotations
concurently.
III. RECENT FACIAL DATABASES
The earlier databases were intended on facial detection for
people identification, the recent databases are tuned more
towards considering the changes in picturing techniques,
facial expressions, and ambiguities due to makeup. Few of
the recent facial databases are 3:
A. Labelled Wikipedia Faces (LWF)
It consists of mined pictures from more than 0.5 million
biographic records from the Wikipedia Living People records.
It consists of 8500 faces from 1500 people. YouTube Faces
Database (YFD) has 3425 videos of 1595 dissimilar people
(2.15 videos per person) with video clips ranging from 48 to
6070 frames. It was designed to give a cluster of videos and
marks for person’s recognition from videos.
B. YouTube Makeup Dataset (YMD)
It is a database which consist of pictures from 151 persons
from YouTube makeup tutorials prior and after precise to
substantial makeup. In this database, four shots were captured
for all persons (two shots prior and two shots after makeup). It
has constant illumination but it illustrates the difficulties in
recognitions of face prior to changes in makeup.
C. Indian Movie Face Database (IMFD)
It is a database which has 34512 pictures from 100 Indian
actors gathered from about 100 videos and cropped to have
changes in illumination, pose, expression, resolution,
occlusions and makeup.
IV. COMPARISON OF FACE DATABASES
The different face databases have been designed for the
analysis of face images when dealing with a single or a
combination of these changes. The different types of face
databases are discussed in the Table1.
Table 1. Different types of face databases 5
Face
Database
Image Type
RGB/Gray
Scale
Image
Size
Types of
conditions
FERET Gray RGB
256 x 384 i, e, p, I/O, t
The yale
face B
Gray Scale
640 x 480 i, p
AR Faces RGB
576 x 768 i, o, t, e
CMU-PIE RGB
640 x 486 i, e, p
The yale
face Gray Scale
320 x 243 i, e
Asian face
database
RGB
640 x 480 p, e, i, o
Indian face
database
RGB
640 x 480 e, p
The different Image changes are shown by p: pose, o:
occlusion, i: illumination, e: expression, t: time delay, I/O:
indoor/outdoor conditions.
The image size, image type and the other specifications
describes about the complexity of face database which in
turn shows the robustness of different algorithms of face
recognition. The different face databases are created to
analyze the effect of changes on the several types of
conditions of an image. AR Faces, FERET, CMU-PIE,
Asian and Indian face database are the most widely used 2D
face image databases. Each database provides a platform to
access the particular challenges of uncontrolled conditions.
For example, CMU-PIE is used for more illumination and
poses changes. FERET gives a good testing platform for
large probe and gallery sets. AR Faces gives the natural
occluded face images. Asian face database has 2D face
pictures of female and male with pose, illumination,
expression occlusion and expressions. The Indian face
database comprises of face pictures with variation in
expression and poses.
