spooky
Originally shared by Wayne Radinsky
A man walks around Amsterdam and feeds the video into the open-source NeuralTalk2 image classification program, which labels every frame of the video by what it thinks is in the video. It gets it right more often than you might think.
John Hardy Developer 
Ozymandias?
LikeLike
very cool
LikeLike
Wow, very cool indeed.
LikeLike
Wrong collection Wayne Radinsky, that’s supposed to be for Aussie political posts.
G+ is not letting me move it right now.
LikeLike
This in Google Glass, make the text spoken, & for a blind person – this would be AWESOME!!
LikeLike
Needs a little bit of tweaking though. You might be wondering why you’re frequently surrounded by a “black & white photo of a … ”
LikeLike
He did film a lot of pavement which is what got misidentified most of the time. On more interesting scenes it got most of it right.
LikeLike
Dino D Fantozzi absolutely. Google Glass wasn’t focused enough on this.
Systems like this could be a huge boon for blind people. For the non-disabled, it’s like adding super powers.
LikeLike
It’s definitely hugely impressive. And no doubt the kind of thing that will improve rapidly. Processing power required must be pretty hefty!
LikeLike