Camera Phones to Interpret Visible World for Blind
iVisit, a company out of Santa Monica, California that specializes in video teleconferencing applications, has teamed up with the National Institutes of Health to turn a mobile phone's camera into an eye for the blind. Using advanced image recognition software to identify what is in front of the phone, the system reads off what it is seeing.
Watch this demo video below of the software in action:
Link.
3 Comments:
But how accurate is it? What if I printed out the words "twenty dollars" on a piece of paper? Would it repeat the phrase "20 dollars"?
Kurzweil has already done this.
KNFB Reader
Either the demonstration is BS or the device has a very limited repertoire. Is it reading the package or simply recognizing a pre-programmed, unique configuration?
For the Frosted Flakes, the device called it "Frosted Flakes Cereal." Cereal wasn't on the package label so it must have been "taught" to recognize that one box. Would it recognize a full-sized box of Frosted Flakes?
The Oats 'N Honey bar was read "Oats anD Honey" and added "granola bar" which was not on the package.
They never scanned the back side of the money, nor did they test out a photocopy of money.
Each item had to be isolated.
My guess is that this limited number of objects was scanned and the device identifies only them. Neat trick, but hardly useful since the universe of all possible objects, configurations, and angles of view is beyond any device's conceivable memory capacity.
We already have facial recognition software that can see through disguises. In Europe they have bills of different denominations in different sizes. It's a cute demo, but I see it going nowhere.
Post a Comment
<< Home