Slashdot Mirror
Google Lens Can Now Recognize a Billion Items (theverge.com)
Google said in a blog post that its AI-powered "Google Lens" camera tool can now recognize over a billion items. When it launched last year, it was only able to detect around 250,000 items. The Verge reports: The expansion comes over a year after the Google Lens' optical character recognition engine has been trained on reading more product labels. By recognizing text, Google Lens thus can put names to the faces of more goods. It has also been fed more data from photos taken by smartphones, so Google says the feature is overall more reliable than before. The 1 billion items figure comes from products available through Google Shopping, so it likely doesn't include more obscure, unshoppable objects, such as a gaming console from the 1990s or the first edition of a rare book. But it covers a huge range of things that could appease someone who's simply just looking up an item they're curious about.
Well, there are 8 billion people in the world, so they're getting there pretty quickly!!
A billion seems like a lot of types of things.
Be precise please.
Where are we going and why are we in a handbasket?
do we notice this is evil
I was waiting so long for google monocle! Its twice as good as muh google glasses.
it only took them 10 years to realize it was twice the waste.
enjoy the spyware!
But can it also detect the elephant in the room?
Or was I blatently tricked into doing away with my privacy when installing that app?
What?
Is this the reverse Google image search or?..
But can it separate the billion things according to the one true taxonomy?
Ydco co
facing resistance from below, the moms have decided to expand the venue for their inspirations, to now include the moms of the amazon, & likely the moms of many other rivers.. talk about anticipating participation? yikes almighty? see you there
Announces Dr Evil sucking the same finger he just had up his bum!
All it takes is to have that software embedded into the Android OS, be always on.
We're so fucked.
How is it that companies like this are free to do whatever they want without consequences?
Probably just a buzz word that sounds cool, but is far less interesting and far less evolved than they want you to think and probably can't be evolved much further.
Maybe in the right conditions with ideal lighting it can get it 35% of the time that it worked, and maybe 7% under normal conditions for certain items.
If it were truly reliable, it'd be in Google self-driving cars detecting stop signs, stop lights puppies, the lady that got run over by a self-driving car and so on.
I'm curious with all these deep learning AI networks that are specializing either in objects, driving, speech etc what the amount of disk space and/or memory does it take? Is all this information stored in a database or in a raw indexed file????
But it can detect a boot stomping on a human face forever
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
Can it recognize my bag in a conveyor belt at an airport, when it is upside down, somewhat dirty, and under somewhat unusual lightign conditions? I didn't think so.
You mean using pictures stolen from smartphones. You remember those pictures your "friend" took a couple years ago? Yeah, parts of YOU are now recognized (and remembered) by the A.I. (Google is working out the details of the monthly subscription service they'll be offering people to 'hide' their least memorable moments from public view...)
It recognizes them....and calls the all a hot dog ( or not a hot dog).
Lens couldn't identify my Christmas cactus but Snapseed had no trouble with it.
Lens thought my Mobius keychain camera was "fudge", as in food.
Not Hotdog?
Hot dog.
Not hot dog.