Vitamin D: Former Palm Execs Revamp Security Video

Vitamin D, a two-year-old Palm spin-off that’s using Palm founder Jeff Hawkins’ new company Numenta‘s natural intelligence technology, is building something a bit unexpected considering its pedigree: better software for security cameras.

Search view MacVitamin D’s first product, launching at the DEMO conference today, uses Numenta’s research on object recognition. The Vitamin D Video software makes processing security camera video simple. Users hook up any webcam or IP-enabled camera to Vitamin D Video (to be available in public beta this fall; pricing TBD), and get a smart and what looks to be extremely usable way to parse long durations of video footage.

Vitamin D Video improves on expensive existing approaches to identifying people in security video. Security camera systems often require special cameras and rely on rules about aspect ratios and motion detection that result in lots of false positives, requiring manual review or continuous fine tuning by technicians to actually process all that video, said Rob Haitani, Vitamin D’s chief product officer, in an interview earlier this week. Instead, Vitamin D apparently recognizes patterns and learns as it goes.

The Vitamin D Video software, available for Macs or PCs, allows users to create rules to quickly parse hours of footage. In a demo Haitani showed me, he took a feed from a fixed camera of a sidewalk corner with a mailbox on it. He narrowed the footage to when people were in the scene. Then he narrowed the footage to only clips in which a person stayed in the vicinity of the mailbox (which he indicated to the program by drawing a box around it) for at least 6 seconds. Turned out there were only three such instances in all the tape — a couple people mailing letters and the mailman. The software demos really well — it identified the clips in question within moments.

Being able to quickly drill down on mountains of footage could have all sorts of applications for discovering what goes on in a place when you’re not there. Vitamin D Video can even understand when a person arrives in a space through a certain door — it’s really cool stuff. Because the software does learn, Haitani said that, now that it knows people, the next things to teach it to recognize will be vehicles, pets, and maybe even specific people. A video demo of what works now is supposed to be made available at this address at some time on Wednesday.

Haitani told us he’s well aware that improving the scan-ability of security camera footage is not a fuzzy feel-good project — but he said he thinks that Vitamin D Video’s ease of use will inspire more casual forms of keeping track of things. Say, a wildlife enthusiast could train the software to recognize when birds fly into his backyard, and save the clips for identifying them.

I’m not sure I buy the new-markets-for-passive-watching argument, but the human tendency to nosiness may prove me wrong. And as footage from hidden cameras gets all the easier to process and cut into meaningful clips, I’m not sure consumers and small businesses are going to go around passing out waivers to everyone who ends up on camera, and then later, perhaps, YouTube. It’s certainly the type of thing celebrities should be praying that stalkers and paparazzi never get ahold of!

Menlo Park, Calif.-based Vitamin D, which has eight employees, has raised $2 million in funding from phonemaker HTC, the CIA’s investment arm In-Q-Tel, and individuals. It hopes to raise another round as it goes to market, Haitani said.

loading

Comments have been disabled for this post