Document Name:	Face Recognition Quick Start up Guide
Document Number:	I-Cube-001.doc
Revision:	A
Date:	19 November 2003
Electronic File:	.......\I-Cube\user manuals
Prepared for:	Development
Author:	Barry T. Dudley (btdudleyAT I-Cube DOT co DOT za)
Configured by:	BTD
Security Classification:	None
Distribution:	Web site (www.i-cube.co.za), CD and Installed with Systems
Distribution Format:	Word (.doc) / PDF file

I-CUBE DataBase Search

The I-CUBE Database Search is designed to simply, easily, and quickly get a fully functional face database search application up and running.

Figure 1 I-Cube DB Search desktop icon

Face Image Comparison

Open the application by double clicking the short cut

on the desktop

“I-CUBE Database Search” or select the application from:

Start, programs, I-Cube, I-CUBE DB Search.

Figure 2 I-Cube DB Search application

2.1.1. Selecting the (White Male) Database

Figure 3 Open an existing database

Select ‘OPEN’ from the File menu or Press the OPEN button on the toolbar.

Figure 4 Selecting the white male database

Select the database to open:

“C:\face recognition\images\male\white\white male.txt”

Figure 5 White male database

The total number of images in the database is listed on the bottom right. Next to the is the actual image number currently selected.

Figure 6 Column names

There are 4 columns:

Image – lists the name and type of image.

Alignment – indicates if the image is aligned (none, poor, low, Medium and high)

Full –

Vector –

2.1.2. Selecting a Subject

Figure 7 Importing a subject

The ‘subject’ is the person who is to be searched for in the database. You can select a subject from the currently opened database by selecting a person and then pressing the ‘Set as Subject’ button or selecting the ‘Set as Subject’ item in the ‘Records’ menu.

Figure 8 Subject directory selection

It the subject is not part of the database, then use the ‘Import Subject’ under the File menu, or press the Import Subject Button, or if the subject search dialog is opened, you can simply drag and drop the image file to set it as the subject.

Figure 9 Subject image selection

Figure 10 Subject image display

Once a subject is set, its face position must be determined. This can be done manually by clicking on the eye positions with the mouse or by pressing the Auto Alignment button on the subject search dialog.

Figure 12 Auto Alignment of the eye positions

Figure 11 Manual eye position selection by clicking and holding

2.1.3. Searching for a Subject

There are different types of searches available. The types of searches available are Standard, Scan, and Hierarchical (and variations of each). All types are available through the main menu under Database, Search or from the subject search dialog from the drop down list. Suggested to use standard search initially until the database size is over 500 000 images.

Figure 13 Search results

Figure 14 Identical image search results

The results are ranked according to confidence, being the degree of match beter the two images. The name of the image is listed on the right. An identical image obtains a value of 10. Use the arrow key down to see the other possible matchs.

Figure 15 Search results ranking

Figure 16 Closest face image search result

Due to changes in lighting, size, colour, orientation, reflections, glasses, hair, expression, etc. it may be necessary to check the top 5-10% of search results (images) in order to confirm that the person is not in the database.

2.2. Capturing own images

The are a number of ways to capture images. In all of them please ensure that the best quality image possible is captured. The lap top has the ability to hold over 2 million images, thereafter a larger hard drive is available. Hence the best quality image is suggested to get the best results out of the I-Cube face recognition system.

The following image files are supported: JPEG (*.jpg), Bitmap (*.BMP), GIF (*.gif), Tiff (*.tif and tiff).

Figure 17 Closest face image search result

When saving the images, ensure they are saved to their own directory, so that each person would have a directory of all images relating to one person.

2.2.1. Analogue image capture (CCTV input)

Open the image capture application.

Figure 18 CCTV camera application

When the person is in the field of view, capture a number of images.

Save the image as a JPEG image to the persons directory. If required edit the image with I-Cube Media Editing application.

Figure 19 Demonstration of the optimal size of the face in the image

Figure 20 Digital image capture

2.2.2. Digital camera input

Ask the person to stand against a light wall.

Take a number of images of the person.

Transfer these images to the persons folder on the lap top.

If required edit the image with I-Cube Media Editing application.

2.2.3. Existing digital images

Any existing images will be added for free to the database for you by I-Cube when the order is placed. Either e-mail or write all images to CD and send to I-Cube at:

ordersAT I-Cube DOT co DOT za

2.3. Creating own/new image database

Open the I-CUBE DB Search application

Figure 21 I-Cube DB Search desktop icon

Figure 22 I-Cube DB Search application

Add images by either drag and dropping images onto the application or by selecting ‘Import Record’ from the file menu. Note, you can drag and drop entire directories.

Once all of the images are added to the database, you must mark the eyes of each person. This can be done manually or automatically.

Figure 23 I-Cube DB Search desktop icon

In order to manually mark the eye positions, make sure the ‘Toggle Alignment Mode’ button is depressed on the tool bar and then left click on each eye in the image. To automatically mark every eye position, select the ‘All Records’ menu item under the Database/Auto Align menu. For a single record, press the Auto Align button on the toolbar, or select Auto Align from the Record menu.

Figure 24 Press and hold to zoom in

Figure 25 Click once on the eye to mark the position of the eye

Marking the eye positions is one of the most time consuming parts of preparing the database. Accurately marking the eyes is essential for good recognition performance.

Figure 26 Press and hold to zoom in

Figure 27 Click once on the eye to mark the position of the eye

Typically you will Auto Align the entire database, then look to see which records were aligned with less than a ‘medium’ confidence. Those leftovers can then be manually aligned or the FaceItDB settings can be re-adjusted and the records can re-aligned with ‘All < Medium’ under the Database/Auto Align menu.

Figure 28 Auto Align all records

Figure 29 Sort records by alignment then manually fix

2.3.1. Create the Face Templates

Figure 30 Create Full Templates

Before you can search a database, you must create the Face Templates. You can not search quickly if you search directly on the images. Each face must have a corresponding template stored in the database. These face templates are about 3.5k in size. In this application, the templates are stored in a single, large file outside of the database. This is called the FID (FaceItData) file. Each face in the database has associated with it a description of where to find its corresponding template. In this case, we store the file offset of the data.

Select ‘Create Full Template’ from the Database menu to create the templates. YOU MUST BE ON-LINE (connected to the Internet) before you do this. Before or after you have created the templates it is a good idea to de-fragment your hard drive. This helps increase search speed. To further increase search speed one may want to create an additional template called a vector template. This is an additional 88 bytes per face. However, on larger databases it can drastically increase the search speed. Select ‘Create Vectors’ from the Database menu to create these vector templates.

IF YOU HAVE ANY PROBLEMS PLEASE E-MAIL SUPPORT

AT I-CUBE (support@I-Cube.co.za)

Smart Cards, RFID, Biometrics and Hand-Held Devices – Advanced Technologies Adopted by Hospitals

Tier I Hospitals Spending Approximately 2% of Their Revenues on IT - New Purchase Plus Maintenance and Upgrade

MUMBAI, India--(BUSINESS WIRE)--Frost & Sullivan recently hosted a two day Healthcare IT Executive MindXchange – “IT Advancements in Healthcare Delivery”, at The Leela, Goa. This high profiled summit addressed the dynamic changes taking place in IT adoption by Indian Healthcare delivery services and focused on the advances that would impact them in the near future.

With the Healthcare Delivery space in India continuing on its aggressive growth path; offering state of the art disease management modalities in established centers, providing world class quality care to the inaccessible geographies of the country, and optimizing the cost of healthcare expenditure have become the prime focus areas for all players in the healthcare delivery space. These focus areas have channeled efforts into leveraging innovative technologies and creating a sustainable competitive advantage for the participants and have consequently brought the role of Information Technology to the fore.

The impact of Information Technology in Healthcare Delivery has been significant not only on the process and enterprise efficiency needed to bring healthcare services closer to people but also in aiding and improving the disease management modalities given the convergence of medicine, devices and technology. This is clearly reflected in the fact that the growth of IT in healthcare sector is expected to record a Compounded Annual Growth Rate (CAGR) of 18% (2006-2010).

“IT adoption in Healthcare Delivery has grown at both process and enterprise level to accomplish a wide range of objectives, chief of which are improvement in process efficiency, enterprise effectiveness and customer satisfaction. The implementation of enterprise applications help in achieving process efficiency, reduction in operating costs, inventory reduction, improved customer service, higher profitability and tap new business opportunities,” says Sandeep Sinha, Program Manager - Health IT and Healthcare Delivery Practice at Frost & Sullivan, South Asia and Middle East. “Today’s integrated applications not only allow hospitals to analyze effectiveness of process but also provide tools for strategy and forecasting to make better business decisions,” he added.

At this summit, various key industry players from large hospitals, Government health departments and major Health IT companies addressed and discussed various issues in IT adoption in hospitals. The event focused on the various healthcare IT standards applicable internationally and the need of similar standards in India, their applicability and the kind of work that is being done towards formulating these standards in India.

There were a few panel discussions which brought forward challenges, best practices and ROI on IT adoption. The expert panels agreed that both management and IT-users in the hospitals are equally responsible for successful implementation of various IT applications. The proper alignment of IT for better delivery of healthcare services could only be achieved with sound awareness about benefits and motivation for new application adoption.

Today, Tier I hospitals are spending approximately 2% of their revenues on IT (new purchase plus maintenance and upgrade). Apart from these, hospitals are also going ahead with the use of some advanced technologies like smart cards, RFID, biometrics and hand-held devices to further improve processes.

The major challenge every hospital faces during IT adoption is in the areas of implementation and maintenance services. Therefore, in Indian healthcare delivery market, areas like training, infrastructure, implementation support and maintenance services to healthcare providers seem to be the biggest upcoming opportunities.

Various major IT players – Microsoft, HP and InterSystems, partnering the event also addressed the issues focusing on connected healthcare services and technologies that will help reduce healthcare costs and create an environment of easy accessibility to healthcare delivery in both public and private sector services.

The other key partners for this event included AGFA Healthcare and Sobha Renaissance. Modern Medicare, Express Healthcare, Medical Buyer and e-Health were the official media partners for this event.

About Frost & Sullivan: Frost & Sullivan, a global growth consulting company, has been partnering with clients to support the development of innovative strategies for more than 45 years. The company's industry expertise integrates growth consulting, growth partnership services and corporate management training to identify and develop opportunities. Frost & Sullivan serves an extensive clientele that includes Global 1000 companies, emerging companies, and the investment community, by providing comprehensive industry coverage that reflects a unique global perspective and combines ongoing analysis of markets, technologies, econometrics, and demographics. For more information, visit