arX

iv:1

205.

4463

v1 [

cs.C

V]

20 M

ay 2

012

1

Hajj And Umrah Dataset For Face Recognition And DetectionSalah A. Aly Moataz M. Abdelwahab

Center of Research Excellence in Hajj and Umrah,Umm Al-Qura University, Makkah, KSA

Abstract—In this note, we define a new face recognitionand face detection dataset, called Hajj and Umrah dataset(HUDA) [1]. The new HUDA dataset presents various imagestaken outside the Holy Masjid El-Harram in Makkah during the2011-2012 Hajj and Umrah seasons.1

Index Terms—Face detection, face recognition, and facialdatabase sets.

I. I NTRODUCTION

Every year millions of muslims arrive to perform the holyrituals of Hajj and Umarh in Makkah, KSA, during theserituals the Saudis authorities save no effort to facilitatethestay of the pilgrims in the kingdom. The greatest challenge isthe huge number of missing persons and unidentified deathsevery year. So, an efficient monitoring system is essential toprovide means of tracking missing and lost individuals andidentifying them in order to take the necessary actions. Tosolve this problem, we describe a new Hajj and Umrah dataset,HUDA, which will be used in the CrowdSensing system fordetecting missing and found people [1], see sample classes ofHUDA in Fig. 1.

Face recognition and detection is one of the most complexapplications in the field of Computer Vision. Developinga computational model of human faces for detection andrecognition, though interesting, can prove to be a very chal-lenging task, due to the complexity of human faces. The rapidprogress in PCs’ speed certainly assists the operation of facedetection and recognition to be done in real-time. For example,recognizing one person’s image among thousands of imagescan be done in a few seconds using current PC’s speed.

One of the most crucial steps in the CrowdSensing systemis to detect the individual faces in an image and run aface recognition algorithm through a database of registeredpilgrims. Therefore, the proposed approach is more focusedon developing a pattern detection algorithm that does notdepend on the three-dimensional complex data of the face, butit depends on the general outer Silhouette that is almost sharedbetween all faces. Our goal is to develop a computationalmodel for face detection and recognition that is reasonablysimple, and acceptably accurate under various conditions oflighting, facial expressions, and background environment.

II. H AJJ AND UMRAH DATASET

In the Hajj and Umrah dataset, HUDA, hundreds of humanimages are taken randomly during 2011-2012 Hajj and Um-

1———————————————————————————Thanks to HajjCoRE, Center of Research Excellence in Hajj and Umrah

at UQU, and NPSTI at KACST, agencies for funding this work.Contact Info: salahaly@uqu.edu.sa

1

2

3

4

5

50

1 2 3 4 5 6

10

20

30

40

60

72

Fig. 1: The new Hajj and Umrah dataset presents variousimages taken outside El-Harram in Makkah [1] 2011-2012Hajj and Umrah seasons.

rah seasons. The dataset contains faces from more than25

countries as shown in Fig. 2. More than six images are takenfor each person in the dataset. Further information about thisdataset can be found in CrowdSensing link [1].

2

Fig. 2: Sample set of Hajj and Umrah Dataset (1)

A. Sample Images of Hajj and Umrah Dataset

The HUDA contains images taken during the 2011-2012Hajj and Umrah seasons of a large number of pilgrims (variedraces and appearances). It contains at least six images for eachindividual, in a varied range of poses, facial expressions (openand closed eyes, smiling and not smiling) and facial details(glasses/ no glasses), and in different lighting conditions andagainst random backgrounds. All images are in full-color JPGformat, see Fig. 2 and‘3.

Fig. 3: Samples of Hajj and Umrah Dataset (2)

B. CrowdSensing System

The CrowdSensing system is established to support the ex-isting efforts to manage the crowds and solve the missing andfound problem during Hajj and Umrah seasons in KSA. Thegoal of this CrowdSensing system is to use techniques fromComputer Vision and Image Processing to develop a portalwebsite for Hajj and Umrah missing and found people [1].The system requires all pilgrims to register their personaldatawhen they plan to perform Hajj or Umrah.

One application of the proposed dataset is the CrowdSens-ing system, which consists of three main components, seeFig. 4:

1) A database of all individuals arriving at the kingdom toperform the holy rituals. This database contains all theirpersonal information along with a personal photo, andit can be updated via our web portal.

2) Advanced monitoring cameras scattered around theGrand Mosque in Makkah, airports, hospitals, and allareas of interest.

3) Our proposed face detection & recognition algorithm isto be used for acquiring faces from images captured bythe monitoring cameras and use them to identify missingand found individuals [1].

III. OTHER FACE DETECTION AND RECOGNITION

DATASETS

Some other face recognition research work can be foundin [4], [5], [7], [8], [10], and face recognition datasets [2],

Data

Collection

Image Search

Engine

Alert

system

Texts - images - videos

Hajj Missing & Found

Fig. 4: HajjMF portal interface website developed by Crowd-sensing.net team. The system consists of data collectors (mo-biles, cameras, PCs), main server, search engine, and analerting system.

[3], [6], [9]. We briefly describe a note about each knownface recognition datataset such as FAFFE, UMIST, ORL, andFERET.

A. ORL Dataset

The AT&T Database of Faces, (formerly ’The ORLDatabase of Faces’), see 5. The database was used in the con-text of a face recognition project carried out in collaborationwith the Speech, Vision and Robotics Group of EngineeringDepartment at Cambridge University. There are ten differentimages of each of40 distinct subjects. For some subjects,the images were taken at different times, varying the lightingand facial expressions. All images were taken against a darkhomogeneous background with the subjects in an upright,frontal position (with tolerance for some side movement). Theimages are in PGM format. The size of each image is92x112

pixels, with256 grey levels per pixel [2].

Fig. 5: Samples of facial images: ORL Dataset

B. Japanese Dataset

The Japanese database consists of10 persons. The imagesper person vary from20 to 23, where facial expressions arevarying [3], see Fig. 6.

3

Fig. 6: Samples of facial images: Japanese Dataset

C. UMIST Dataset

The Sheffield (previously UMIST) face database consists of564 images of20 individuals (mixed race/gender/appearance).Images are numbered consecutively as they were taken. Theimages are all in PGM format, approximately220x220 pixelswith 256-bit grey-scale [6], see Fig. 7.

Fig. 7: Samples of facial images: UMIST dataset

D. FERET

The Face Recognition Technology (FERET) dataset is oneof the known face recognition datasets that was collectedbetween 1993 and 1996 and was developed by DARPA [9],see Fig. 8.

Fig. 8: Samples of facial images: FERET dataset

IV. D ISCUSSION ANDCONCLUSION

In this note, we have described a new Hajj and Umrahdataset, HUDA, for face recognition and face detection. Dur-ing the huge overcrowds that occur in the two holy cities

of Makkah and Madina in KSA every year (more than5millions people observe Hajj for at least10 days), in additionto five millions people observe Umrah throughout the year.The proposed CrowdSensing system can be used to identifymissing, found, and dead parsons. Further details about thissystem can be found in the project website [1].

TABLE I: Information about ORL, UMIIST, Japanese,FERET, HUDA datasets

dataset # persons # images

ORL 40 persons, Each has 10 imagesUMIST 20 persons, vary from 19 to 48JAFFE 10 persons, vary from 20 to 23FERET too many manyHUDA too many each has at least 6

ACKNOWLEDGMENTS

This research is supported by the Center of ResearchExcellence in Hajj and Omrah (HajjCoRE) at Umm Al-Qura University in Makkah, Kingdom of Saudi Arabia, undera project number P1127, entitled ”Crowd-Sensing System:Crowd Estimation, Safety, and Management”, 2012. Also,this work is funded by a grant number 11-nan1707-10 fromthe Long-Term National Plan for Science, Technology andInnovation (LT-NPSTI), the King Abdulaziz City for Scienceand Technology (KACST), Kingdom of Saudi Arabia. Wethank the Science and Technology Unit at Umm A-QuraUniversity for their continued logistics support.

REFERENCES

[1] www.crowdsensing.net. Copyright, 2011.[2] AT&T Laboratories Cambridge. Database of faces,

(formerly ’the orl database of faces’), available:http://www.cl.cam.ac.uk/research/dtg/attarchive/facedatabase.html.

[3] Kyushu University Psychology Department. The japanesefemale facialexpression (jaffe) database, available: http://www.kasrl.org/jaffe.html.

[4] Pallavi Vaidya Divya Jyoti, Aman Chadha and M. Mani Roja.A robust,low-cost approach to face detection and face recognition.InternationalJournal of Digital Image Processing, 2011.

[5] D. Jyoti, A. Chadha, P. Vaidya, and M. Mani Roja. A robust,low-cost approach to face detection and face recognition.CiiT InternationalJournal of Digital Image Processing,, 15(10), October, 2011.

[6] The University of Sheffield. The umist face database, available:http://www.sheffield.ac.uk/eee/research/iel/research/face. 1998.

[7] J. Malik S. Belongie and J. Puzicha. Shape matching and objectrecognition using shape contexts.IEEE Transactions on Pattern Analysisand Machine Intelligence, April 2002.

[8] D. Garg Singh. Soft computing.Allied Publishers, 2005.[9] Viola and Jones. Rapid object detection using boosted cascade of simple

features.Computer Vision and Pattern Recognition, 2001.[10] J. Yang, D. Zhang, and A.F. Frangi. Two-dimensional pca: A new

approach to appearance-based face representation and recognition. IEEETransaction on Pattern Analysis and Machine Intelligence, 26(1):131–137, 2004.

4

1 2 3 4 5 6 7

1

2

3

4

5

6

7

8

9

10

11

Fig. 9: Hajj and Umrah dataset (HUDA) for face detection and recognition