Quantcast

Object Tracking Algorithm Learns From Mistakes [Video]

By Wesley Fenlon

The Predator algorithm can track faces and other objects and filters out excess visual data to improve its accuracy.

Face tracking technology is nothing new, but a Czech student at the University of Surrey in the UK has developed one hell of a camera system dubbed the Predator. More specifically, it’s an object tracking algorithm that hones in on a selected object in a video and filters out the surrounding visual data to pinpoint its target. It learns what the object looks like as it moves and rotates on camera, so changes in orientation--or even leaving the screen--won’t throw the Predator off.


 
  
We don't know what's more impressive here: the variety of potential uses for Predator's smart algorithm of the fact that the system learns from its mistakes to track objects with more acuity. The brainchild of the operation, student Zdenek Kalal, won a "Technology Everywhere" award for his work. Not a bad endorsement for a Ph.D. thesis.
 
The project's code is available on the University's website.