The annual massive techfest that is CES in Las Vegas has not long finished and winner of the accessibility category, new to 2017, is a revolutionary app that uses AI to help recognise the world for people with low or no vision.
AIPOLY Vision – giving the world a voice
On the face of it AIPOLY is like many other apps that can recognise objects using a smartphone’s camera, and then speak out what it ‘sees’ to give blind people, like myself, rich information about what is around us. Here is a short video of AIPOLY in action.
Not just another object recognition app
I’ve previously written and demonstrated about the magic that is Talking Goggles and still use apps such as this all the time to help recognise products on shelves in a supermarket or read the text of a menu in a restaurant.
All these apps, however, need an internet connection to work. The brains of the system aren’t found on the phone but rather up in the cloud somewhere. This is fine so long as you have a good connection, but as soon as you’re out of signal you’re in the dark.
AIPOLY is the first app that stores the smarts within the app so no internet is needed. What’s more it actually learns from what it sees and, when a connection is available, it sends its newly recognised objects up to the cloud to make future updates of the app even smarter.
For these reasons AIPOLY snagged best of class in the CES innovation awards.
Lightning-fast help wherever you are
AIPOLY is quick. It analyses the camera image three times a second and, as we see in the video above, it’s fast to feed back what it thinks it sees. It doesn’t always get it right but its right enough of the time to make it incredibly useful - and it’s learning.
Combining such a useful app with a head-mounted camera would make AIPOLY a winner for me. Having a camera with this kind of intelligent object and text recognition, observing and analysing whatever it is I’m looking at, is a dream I’ve had ever since the advent of Google Glass ushered in a new era of raised awareness of our surroundings.
And of course this smart object recognition doesn’t stop with helping the blind to see – in fact we’re just lucky beneficiaries of the massive mainstream initiative to develop more and more intelligent machine learning to better understand big data. Good luck to them.