Click here to Skip to main content
Rate this: bad
good
Please Sign up or sign in to vote.
See more: C++ C Java Android AI OpenCV
Well
 
For a college project, we have to create a Face recognition module on android. So i went through certain tuts and info. on internet and also got lots of information about PCA,Eigen Faces, Elastic bunch graph etc.
 
Now the point is i have already done my face detection portion, and for face recognition i thought of using a different approach. so instead of comparing test images to eigen faces in database, is it possible to create a face map on captured image and calculate the relative distance of various features of human face. And from this measurements a long alphanumeric key will be generated, something like this:
 
EY1.8888NO4.1111LP0.8892CN8.64324ER
 
And only this key is sent to the server for comparison. and i guess it should generate the same key for the same person every time.
 
So i desperately need a help, on how i can do this, which algo can i use and what sort of precautions i need to take and also your views on my concept (i.e. its good?? or just crap).
So please help me.
Thank Yoou
Posted 10-Jul-13 12:33pm
Comments
enhzflep at 10-Jul-13 18:23pm
   
Your method is quite dependant on the resolution of the image. Further to that, it sounds like you would like to perform the recognition step upon multiple features in the face, rather than the face as a whole. It sounds like it would be susceptible to speed and accuracy issues.
 
It would also depend upon orientation and the ability to discern orientation and make corrective calculations for this, such that accurate feature to feature dimensions may be obtained. Sure, the idea sounds attractive, in that a fairly small quantity of data (the key) would identify each face.
 
Sounds like an interesting approach. Not one I'd personally bet on, but interesting all the same. :thumbs-up:
The_Inventor at 11-Jul-13 2:26am
   
I'm betting that "The Machine" on CBS2POI, knows how to do that already. Multiple samples, laser scanned, etc. Correct scale function, from multiple 2D, as in 'what size is it really if I had a tape measure to measure that 'huge' nose' is difficult at best. There was a program that would use two 2D images of the same building, imaged from two different angles, with a carpenter's square in both images for calibration. The finished results were a render ready 3D models that were surfaced with the images in the photos as the texture.
The_Inventor at 13-Jul-13 22:17pm
   
I am sure you already have the algorithm for your output. The key would change, because the person changes, daily. Recognition systems are somewhat complex, but all current commercial ones use a 2D image method, against a known base image in an accessible database, much like the feds have in place now. They update your file every time you fly. Height, weight, internal and external structures, temperature, etc. all the data is collected and stored.

1 solution

Rate this: bad
good
Please Sign up or sign in to vote.

Solution 1

Multiple face detection and recognition in real time[^]
 
use this sample article,
 

for comparison ,
 
get hex values for detected images. save it, and compare it for further images.
  Permalink  

This content, along with any associated source code and files, is licensed under The Code Project Open License (CPOL)

  Print Answers RSS
0 OriginalGriff 584
1 Maciej Los 275
2 DamithSL 233
3 Sergey Alexandrovich Kryukov 209
4 BillWoodruff 200
0 OriginalGriff 6,803
1 Sergey Alexandrovich Kryukov 6,377
2 DamithSL 5,421
3 Manas Bhardwaj 4,841
4 Maciej Los 4,330


Advertise | Privacy | Mobile
Web04 | 2.8.1411023.1 | Last Updated 2 Feb 2014
Copyright © CodeProject, 1999-2014
All Rights Reserved. Terms of Service
Layout: fixed | fluid

CodeProject, 503-250 Ferrand Drive Toronto Ontario, M3C 3G8 Canada +1 416-849-8900 x 100