Sunday, 2 October 2011

Face recognition ..continued


This technology is called Veriface technology…the basic two steps are as above. The snapshots are black and white.

Then there is live detection. This is done by monitoring the eyeball movement and blinking of eyes of the person in front of the webcam. This is done to ensure that someone does not use the photograph of a person to break into his computer.

A log of the recent logins is maintained alongwith the snapshot of the face used for login-keeping the user aware of the people who might have tried to break into his system.

This technology might excite us all and we might be ambitious enough to call this invincible but if you just skimmed through the lines ,please go back to [1]. It says the image is digitized and then converted to 3-D counter part . Face-recognition algorithms cannot tell the difference between a digitized image and a real face. Because the algorithms, in effect, process digital information sent via the camera, it is possible to trick the software with an image of a registered user of a system.

So what if we use digitized pic of a person to break into his system.Normal photograph isn’t digitized plus we can’t move our eyes and blink it.So I tried to break into this system by recording a clip of mine in a digicam wherein I blink my eyes normally and then kept it in front of this webcam and Viola!! It recognised my face.I had breached lenovo’s Veriface fortress.

Here's the link of the video where i break through lenovo's vary face . i would realy appreciate if u all have a look at this as we worked hard on it.


Sorry for the language inside the pic, my friend who owned this lappy was French.

This uses a little more complex algorithm to recognise collects nearly 5000 data points compared to lenovo’s 2000. It supports a face with hat and glasses as there are more data points to be used as a replacement.

However this too was forged with some effort..this uses face movement as a tool for checking liveness.


This used the most effective algorithm…everything was different over here…this could not be forged at all effort to find out the details of the algorithm from both the vendors and the users vent in vain. But as far as I know ,this should be collecting more data points in the face than any other .But it has the least user convenience and more emphasis is given to security.

My piece of code

A piece of code written by me in python and by importing the OpenCV module. Open CV is the language used to control the camera efficiently .

Courtesy -Ankit Sarkar ,4th year IIIT-D

import sys

from CVtypes import cv # the open cv module ie the open camera vision module that needs to be imported to make this code run

def detectface(image): # a function I maded to detect the image

image_size = cv.GetSize(image)

# this is to create the grayscale version of the image-makes it easy to recognise the skin tones and colour temperature

grayscale = cv.CreateImage(image_size, 8, 1)

cv.CvtColor(image, grayscale, cv.BGR2GRAY)

# create storage for the features of the face and the pointer is placed at location specified-0th byte

storage = cv.CreateMemStorage(0)


# equalize histogram for the grayscal image

cv.EqualizeHist(grayscale, grayscale)

# detect objects using the haar classifier..that is the tool which can make frames to detect faces more efficiently.

cascade = cv.LoadHaarClassifierCascade('haarcascade_frontalface_alt.xml', cv.Size(1,1))

faces = cv.HaarDetectObjects(grayscale, cascade, storage, 1.2, 2) cv.HAAR_DO_CANNY_PRUNING, cv.Size(50, 50))

I didn’t know how to go ahead to extract the features,this just creates the classifier which is the main part. Hopefully ,within some time I am able to perfect my OpenCV skill.

A tryst with comics

With permission from my partner Gandharv Kapoor here is what he has to feel on face recognition technology in laptops.

From a serious side, the funda in face recognition is –security ,comfort and cost-choose any 2,you can’t have all three.

Keeping this in mind-Asus was the best as it provided the best comfort to safety ratio.


Thanks first to Amarjeet sir for this unique concept and opportunity, then to all the patient viewers of my post, to Ankit Sarkar for his great contribution in the piece of code ,my friend Richard for his Asus lappy, all the Googlers for the amazing search engine they created and lastly to Steve jobs for his guest appearance . This blog was dedicated to him, after all apple is going to use face recognition –iFace in its iOS 5.Hope I helped its cause.

Thanks a lot



Sriram Ramesh(2011113)

Gandharv Kapoor(2011047)

No comments:

Post a Comment