Quick, Draw!
-
https://quickdraw.withgoogle.com/#
This is a game built with machine learning. You draw, and a neural network tries to guess what you’re drawing. Of course, it doesn’t always work. But the more you play with it, the more it will learn.
This is fun for killing a few minutes, and of course to impress other people by sharing your mad art skillzzz.
-
-
-
There were a few times when my browser lagged so I had a straight line instead of whatever I actually drew, but it still got them ok
-
@loopback0 Your wildlife drawing skills are amazing.
-
-
-
-
-
It's pretty good at drawing premature conclusions.
-
-
Third match
-
Table...
This is totally a complete hourglass.
-
I can't believe it didn't recognize my sleeping bag and my Trump ET.
-
-
I'm proud of this one.
-
Chandelier was surprisingly simple
-
Come on
-
Nice easy round...
-
Book was Great Wall of China until I drew two lines of text.
Good job with the Eiffel Tower.
-
Doesn't recognise cake:
But doesn't even let me finish drawing keyboard:
-
@hungrier said in Quick, Draw!:
here were a few times when my browser lagged so I had a straight line instead of whatever I actually drew, but it still got them ok
It's stupidly lagging whenever their wish-it-was-Siri starts talking. And once she starts, she doesn't stop anymore. Really messes up drawing, which is ridiculous considering a 286 could do this.
-
Novelty wore off for me, I'm going to call it a day.
-
-
-
-
@topspin said in Quick, Draw!:
@hungrier said in Quick, Draw!:
here were a few times when my browser lagged so I had a straight line instead of whatever I actually drew, but it still got them ok
It's stupidly lagging whenever their wish-it-was-Siri starts talking. And once she starts, she doesn't stop anymore. Really messes up drawing, which is ridiculous considering a 286 could do this.
Just drew something, it couldn't figure it out because it turned everything into straight lines, but was still guessing after I was finished and already drawing the next thing. Drew slower for that one, it got it correct but still had to finish saying its 10 wrong guesses it had accumulated on the way.
So basically the screen tells me it correctly guessed "cup" and Siri is still saying "I'm guessing line".
-
@Zecc said in Quick, Draw!:
I'm going to call it a day.
I lied. I gave it another shot.
What is it with keyboards?
-
I only drew the additional appendage on the shark after it stubbornly failed to recognize it
-
@hungrier LOL @ Zebra
I didn't have time to draw the inside of the thermometer, so it failed to recognize hot dog:
Now I'm leaving. Have fun guys.
-
-
@topspin said in Quick, Draw!:
@hungrier said in Quick, Draw!:
here were a few times when my browser lagged so I had a straight line instead of whatever I actually drew, but it still got them ok
It's stupidly lagging whenever their wish-it-was-Siri starts talking. And once she starts, she doesn't stop anymore. Really messes up drawing, which is ridiculous considering a 286 could do this.
Oh Christ, I just tried it on my laptop and it's like that.
Was fine on my desktop though.
-
-
@hungrier said in Quick, Draw!:
-
-
@hungrier I see what you did there.
-
-
-
I was asked to draw a train. I drew a train and it thought it was either a toothbrush or a screwdriver. Turns out that everyone else asked to draw a train drew a steam locomotive.
-
-
Um... okay, I'll have a crack at it.
-
-
(check the URL hiding behind the title shown by the embed)
In 2018 Google open-sourced the Quickdraw data set. “The world's largest doodling data set”. The set consists of 345 categories and over 15 million drawings. For obvious reasons the data set was missing a few specific categories that people enjoy drawing.
Anyway I think Google’s QUALITY code might have pretended to work despite depending pretty crucially on a few scripts that weren’t allowed to load…
(hedgehog looks shit because the button that looks like an eraser actually clears the whole thing. With just a few seconds to go…)
-
@kazitor I followed instructions: I drew a vulva.
Or an onion, as it was repeatedly identified.
-
-
-
Bananas are a lot like the great wall of China.
Fear my animal drawing skills.
Sure...
-
It's funny the way it develops biases - I drew a typical modern broom with the bar-shaped head, and it didn't get it. When I compared with other drawings, everyone had drawn an old-fashioned with a bundle of sticks, like a witch's broom. It didn't guess my pig which I thought I did relatively well, but I drew it in profile, and it turns out everyone else draws pigs face on. It wouldn't recognise a belt that wasn't drawn fastened. (It also wouldn't guess "pants" when I drew underwear, but I knew that was going to happen.)
I tend to stick to my guns if I've started drawing something one way, and just continue to embellish it until it's guessed or I run out of time. I wonder how many people start over with a different approach if it doesn't get it quickly, and whether that results in reinforcing the net's preconceptions.
-
@CarrieVS said in Quick, Draw!:
It's funny the way it develops biases - I drew a typical modern broom with the bar-shaped head, and it didn't get it. When I compared with other drawings, everyone had drawn an old-fashioned with a bundle of sticks
"Flipping" image recognition neural nets is an important for understanding how, exactly, they're recognizing objects. If you want a neural net to recognize dumbbells, for example, you might send it a bunch of images of people doing arm curls with dumbbells. Ideally, the nets will just notice the weights. But sometimes, as in the case of one Google neural net, they pick up too much, and think that all dumbbells have to have muscly arms attached:
-