Face Recognition Algorithms – time for some rethinking!

Face recognition - http://www.face-rec.org/

I think it’s high time we admit it; despite all the work that has been done (and all the complex maths used) on face recognition systems, the performance of these algorithms and systems are abysmal. They certainly do well in lab conditions (error rates of maybe 1 in 50) but in real life conditions, they suck!!!

From the approximate figure mentionned above (1 in 50), that would seem good but it means that if 1000 people were to pass by, we would have 20 errors and in 10000, 200 errors and this is very roughly what we will have to face in real life conditions. Yet, this is not to mention variation in lighting, angle of faces,…

So the question that is to be asked is: are we proceeding the right way? Are our algorithms basically flawed? Are our basic assumptions about recognition wrong?

I think we ought to do a total rethinking of way in which we have been proceeding and try to see what is it that humans use to recognise faces. Human seem to rely on many things like colour of the skin, some oddities of a specific face,… to perform instantaneous recognition.

The other side to it is are we being fair in requesting a 100% recognition rate from computers based only on faces. Is that feasible? Do humans rely only on faces to recognise a person or is there an array of things – “I thought it was Jack but he’s way too tall to be Jack.” Isn’t that something we say a lot?

I’ve put up a lot of questions here coz I sincerely think that we need to rethink our algorithms. It simply can’t be that we’ve been doing research for about 20 years now and our results are sooooo bad!!!

6 responses to “Face Recognition Algorithms – time for some rethinking!

  1. Pascal,
    You are so right on. FR sucks. My company started in 2003 with some brainy science in 3D computational anatomy….a history from recognizing the details of the brain. We adapated it to face. We are converting 2D photos to 3D geometries and thus have so much more info to work with including lighting conditions and head rotation, which we normalize and synthesize. The net is we are attaining what we might call next generation results. We’re a fledgling company, funded by DARPA, and ready to rock the world.

    Thanks for your astute observation!
    Best
    Paul Schuepp, Pres CEO of http://www.animetrics.com

  2. Had two students who worked on something similar for a completely different project. They had to generate 3D faces from 2D photos. We found that just taking photos and mapping them on some generic 3D face models gave quite good results (though not perfect). I wouldn’t be surprised if such a technique could be used to recognise faces. It would de facto solve the head rotation problems that other “2D” based algo struggle to cope with.

    Also, modifying light factors is very easy in 3D and so once a face is obtained, all sorts of lighting effects can be simulated. Furthermore with 3D software, we can easily manipulate the underlying 3D mesh so that with some tweaking, we might no longer need to capture faces with varying expressions, deformation of the mesh (through facial animation might hypothetically do the trick).

    Right now face databases need to record faces with varying illumination and pose!

  3. I am currently my final year project in the field of face recognition; more precisely, the simulation of a face recognition based authentication system. I have to say I totally agree.

    I came across a website about a 3D face recognition project while doing my research. The system is able to construct a 3D face model of a subject using only two images of the subject taken at different angles. The model then allows the simulation of various head poses. The people working on this claim the results are very promising! Sadly, I can’t seem to find the link to the website…

  4. Incorporating 3D techniques into face recognition seems indeed to be the new hype and logically it should solve “pose” related problems – a lot of research is going in this direction these days.

    Yet, I hope that they deliver the goods and that these hyped 3D techniques don’t just stay in the phase of promises!

    The future will tell but I still think that we need to rethink everything and analyse why humans do so well at recognising faces.

  5. Just like you said, I don’t believe our recognition abilities stick to faces only. We can even recognise people while looking at them from the back, can’t we? It’s more of a global thing.

    But then, our brain is a zillion times more powerful than any computer, and we’ve been learning to identify people we see since our youngest age.

    Isn’t that asking a bit too much from the machines?

  6. Y have people stopped contributing to this section just when I have spent hours learning OpenCV and all its pointers. Well hope they reply bak. Yeah the blog’s master was my computer vision lecturer. Thought I had say a ‘hi’ and he could help me with some little pointers in OpenCV. Wat say?

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s