Revolutionary Object-Tracking Video Software Released as Open Source

This young man’s software is causing a bit of a stir.

He’s invented code that doesn’t simply recognize faces, but can track them as they move around. Not just faces, either. Almost any object you define on screen can be monitored, tracked, and memorized for future use.

Zdenek Kalal‘s pioneering work in object tracking has resulted in a system he calls Predator, which he’s just released to the world under an open source license. It’s also won him an award.

In short, Predator can be shown any object – face and fingertips, yes, but also pretty much anything you can imagine – and then track it in a video stream.

It learns as it tracks. It watches for movements in the object, and works out what it looks like from various angles and in different lighting conditions. It’s astonishing stuff.

That crashing noise you can hear is the emails landing in Zdenek Kalal’s inbox. Since announcing this to the world, he’s been inundated with hundreds of emails from people wanting to put his technology to use. Earlier today, he released version 1.0 of his software under the GPL open source licence. The software will run on many common personal computers right now, and could perhaps be adapted for use on cell phones too.

Where will we see Zdenek’s software in the future? Good question. In computer interfaces? Possibly. In police and security stations, monitoring individuals under video surveillance? Sounds plausible. Those and a thousand other possibilities beckon.

Kalal’s PhD supervisor said Predator could go far: “It can also be used in other areas such as human computer interactions, augmented reality, video post-production, motion capture, visual surveillance, robot navigation and assisted driving.”

Remember the name Zdenek Kalal – I have a feeling you’ll be seeing it again one day.

Related Topics: object tracking, technology, UK, Web Video, Zdenek Kalal, Accessories & Peripherals, Gadgets
  • http://kidseyemd.wordpress.com kidseyemd

    This is phenomenal. It has vast, incredible, useful implications.

  • http://manwhojaped.wordpress.com manwhojaped

    Sorry…
    This is nothing new. Maybe new to you, but 20 yrs old tech to me. This is an example of a class of (visual) pattern recognition algorithms. These algos were first used in analyzing moon photos by NASA over 20 yrs ago. My company and others (again – 20 yrs ago) used similar algos as position feedback for robotic systems. Such algos can have many properties: scale invariance (how well does the recognition perform if the object is moved away from the camera), rotation invariance, lighting invariance, etc. If you look closely (I did) this algo does not seem to be rotationally invariant as it loses lock when the hand is rotated.
    Worth a PhD? Hmmm…

  • http://aaronwetzler.wordpress.com aaronwetzler

    @manwhojaped: Really? Would you like to name them or point us to some reference to them? If theyve been working for the last 20 years they must be pretty good.

blog comments powered by Disqus