Back to Tech Blog homepage
Working from home on Friday, I noted 17 cars, 10 people and a squirrel come and go in our small cul-de-sac.
Not that I spent most of the time looking out of the window - I was just trying out new motion-detecting webcam software from Vitamin D, inspired by the way humans recognise patterns. It has been released in a free public beta from Monday.
There’s a pattern to Vitamin D’s development as a company as well. Its three founders were all senior executives at the Palm and Handspring PDA and phone companies.
Jeff Hawkins, Palm’s founder, had a passion for figuring out how the human brain works and set up a neuroscience institute, where he worked out his theory of hierarchical temporal memory and wrote a book on intelligence.
Three years ago, he started a company called Numenta aimed at developing a software platform based on his theory.
Greg Shirai, Rob Haitani and Celeste Baranksi decided they would like to build applications on the platform and started Vitamin D.
Surveillance and security software seemed the most obvious initial application for them, given the pattern recognition capabilities of Numenta’s technology and the rather inefficient existing tools for motion detection, which are based on rules relating to the number of pixels changing in a picture.
The increased accuracy of Vitamin D Video eliminates many of the false alerts associated with motion-detection software.
In my case, it accurately drew yellow boxes around people it spotted and green boxes around other objects, like the cars and squirrel. These movements were listed as clips of time-stamped video that I could easily scan down and watch.
I did notice one false positive when a breeze developed and the software clipped some swaying flowers, while it missed the squirrel a couple of times when my driveway was partly shadowed, although a higher resolution webcam might have improved things.
Vitamin D says testers of the software had found it saved them a tremendous amount of time - instead of going through hours of footage or clips with many false alerts, they could see an accurate list of relevant events. The settings can also be adjusted to focus on just one part of the picture, like a post box or storeroom door, while filters can show footage containing just one type of object.
“We believe we currently live in the Dark Ages of computer video - computers are dependent on humans to tag videos in order to do anything interesting with them,” says Rob Haitani.
He says the beauty of Numenta’s software is that it is constantly learning and improving its pattern recognition, so more sophisticated recognition of objects will be possible in future. Applications will go beyond video, but imagine a camera phone that can recognise a car in its viewfinder and present all the relevant information about the model and features in a version of augmented reality.
In the meantime, expect to see Vitamin D software bundled with future webcams and enjoy trying it out with your own webcam in the beta version - it’s turning me into the neighbourhood snoop!
David Gelles, Joseph Menn, Chris Nuttall and Richard Waters in the FT's San Francisco bureau upload their views - plus tech insights from writers in New York, London and Tokyo
Richard Waters
Chris Nuttall
David Gelles
Maija Palmer
Joseph Menn
Robin Kwong
Tim Bradshaw
The latest gadgets and gizmos, reviewed by Jonathan Margolis in How To Spend It.
Paul Taylor, the FT’s personal technology expert, answers your gadgetry questions