Intro to CV Dazzle

Posted on March 21, 2010, in ITP, Research, Spring 2010, Thesis

One day, instead of asking me how I’m feeling, you’ll ask your camera. It will look at my features, compare me to other people and decide that I’m working on a thesis, in need of sleep and desperately need to get some sunshine. You’ll be able to link this image to my online social network profiles and semantically analyze my status updates and online activity to confirm your results or maybe look at my online calendar to see what I’m doing next. You’ll still say “what’s up” except it wont’ be a question. You just say it so that my device can recognize your voice and then find out what you’re doing, who you’ve been dating , and if your schedule is open tonight. Our PDAs will sync up and suggest a place to go later tonight based on local Yelp reviews, where we’ve been before, the last time we had drinks and when we’re both free.

Most all of these systems exist today. Some of them are built on large infrastructures that are not yet mainstream, like expression recognition and vocal-identity recognition, but even entry level point-and-shoot cameras like the Sony Cybershot include special features like face detection and a “Smile Shutter mode”. Use your computer or mobile device and you move to the next level of sophistication. With more advanced computer vision tools you can play fun games like detecting whether someone is glad or sad, if they want an ice cream cone, their gender, approximate age, or your celebrity doppelganger (for a list of software demos, check the bottom of this post).

But personal computer software is just the beginning. With more money, comes better toys like biometrics that fuse facial, vocal, gait and ocular recognition into government grade identity tracking solutions. Among the most popular of these technologies are facial detection at mid-distances and occular-recognition at close range. Gait detection or Human Identification at a Distance (HumanID), which can track people based on their cadence and body movement, has been a sought after addition to the suite, especially since 911. This gives computers the ability to identity individuals at long range distances (the Information Awarness Offices’ goal was up to 500ft).

Being the most popular means you get all the development money. And given the large demand for facial recognition technologies for security, entertainment and artistic applications, the field has developed rapidly. OpenCV, an open source library for computer vision developed by Intel, has several freely available versions for various programming languages including Java, Processing, OpenFrameworks, ActionScript, and Python among others. Variations of the OpenCV framework can be seen in online advertisements like the Transformers AutoBots campaign and in physical environments like the successful Amnesty Internationalinteractive face-tracking ad that responds to people looking at it.

Beyond the freely available software, many companies develop their own proprietary, more advanced solutions. Most use 2D cameras, but given the recent introduction of 3D digital imaging technology in system such as Microsoft’s Project Natal, I’d expect to see more uses of this. As a result, CV algorithms will become much more accurate in the next few years.

My thesis at ITP, is to research and develop privacy enhancing counter technology. The aim of my thesis is not to aid criminals, but since artists sometimes look like criminals and vice versa, it is important to protect individual privacy for everyone. Since many companies offer free demos of their products, I will use these to test my methods against theirs. The bulk of CV work is fascinating and useful, but the underbelly is, in a way, like a silent paparazzi. In a recent article in the New York Times, In Bid to Sway Sales, Cameras Track Shoppers, Stephanie Rosenbloom mentions that “the most basic surveillance has been around for a few years”, but now “many think they have just begun to tap its potential. I think she is right and as digital cameras and monitoring software becoming cheaper and more intelligent, the question of who owns what data will increasingly become a mainstream issue. In the online world, its certainly been getting a lot of “Buzz”.

What will these forms look like and how well will they integrate into our cultural expectations of body decoration while still being able to function as face detection blocking devices? How can hats, sunglasses, makeup, earrings, necklaces or other accessories be modified to become functional and decorative? These are the topics that I’ll be exploring in thesis on CV Dazzle.

Facial Recognition Software Companies

Associations

  • The Covenant Consortium – Biometric facial recognition and identity tracking. Dallas, TX
  • IBIA – Biometric facial recognition. Washington, DC

Computer Vision Software: Face Recognition and Object Detection

A List of Software Demos

Recent Articles about Surveillance, Cameras, and Computer Vision

Further Reading on Software

Further Reading on Applications

 Trackback
 
All content © 2003-2010 Adam Harvey All Rights Reserved | Contact