With VizWiz, the blind can take a picture, ask a question, and get an answer back from a real person in seconds.
What’s the News: With the web as their eyes, the blind will able to read menus, identify canned foods, and tell whether that park has any free benches without having to walk over. That’s the vision of a team of computer scientists behind an iPhone app called VizWiz, which lets people take a photo of whatever’s perplexing them, record a question like “What denomination is this bill?” and send it off to real people online who’ll respond in a matter of seconds with “That’s a 20.”
How the Heck:
- Blind people have workarounds for the kinds of tasks the sighted use their eyes for—folding dollar bills in certain shapes, keeping the cans of tomatoes separate from the cans of beans, and so on—but these measures often require the input of a sighted person at some point, and they’re not very efficient. An app like this would give the blind more independence.
- Many simple tasks, like reading an address off a letter, are phenomenally difficult for computer intelligences. So the scientists are working with Amazon’s Mechanical Turk, a system that employs people around the world to do odd jobs computers can’t, like choosing the best picture for a product website or reading signs in photographs.
- Then, to solve the problem of speed, the team wrote a program called quikTurkit that works to recruit people even before the question is sent, so there’s always someone on hand to answer. In the latest version of VizWiz, the average turnaround time on a question was 27 seconds. Not bad.
The Future Holds: VizWiz, which is being tested by teams of blind volunteers, hasn’t left the lab yet. But the volunteers are fans: it would be “very useful,” one said (via New Scientist), “because I get so frustrated when I need sighted help and no one is there.” Though an in-depth study [pdf] on VizWiz was released last year, there’s no word yet from the scientists on when this will hit the market. Soon, one hopes.
(via New Scientist)
Image credit: Rochester Human Computer Interaction Group