[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/ic/ - Artwork/Critique

View post   
File: 828 KB, 1180x1520, Proko neural net.png [View same] [iqdb] [saucenao] [google]
3362972 No.3362972 [Reply] [Original]

oh this doesn't seem good

>> No.3362974

>>3362972
was this an email sent to only artists from /ic/?

>> No.3362979

>>3362972
>literally send bad artwork
Don't worry, I got this.

>> No.3362985

He's basically tired of looking at /beg/ tier work

>> No.3362987

This somehow needs to be installed into art university curriculum so students can actually learn without their crab teachers praising garbage post modern "style"

>> No.3362991
File: 71 KB, 557x551, kangaroo.png [View same] [iqdb] [saucenao] [google]
3362991

>>3362972
I did my part.

>> No.3363807

>>3362991
But that's tier 5 you dingus.

>> No.3363832

>>3362985
That's kind of sad, I never thought about the random emails he gets from many fans all over the world, asking for critique on their /beg/ tier drawings. I wonder if he actually responds to all of them. Did any of you ever submit anything after watching his lessons? he used to encourage it

>> No.3363867
File: 86 KB, 499x679, tumblr_p01vfmcAP21w49lylo1_500.jpg [View same] [iqdb] [saucenao] [google]
3363867

>>3362972
>trusting algorithms to competently rate art

>> No.3363885

>>3362972

I'm doing a CS student and what he is trying to do is impossible.

Neural networks aren't magic and unless he made some fundamental breakthrough then meh.

Google deepmind is probably at the top of the field with this and even they couldn't manage this.

Even looking at the picture you have a a bunch of faces, one that fills up the whole canvas. One that is a body shot. One that muscular.

What is Proko even going to do with that crap. Feed it into a CNN?

https://deepmind.com/blog/learning-to-generate-images/
This is what deepmind people can do. I feel they actually are stuck as they have not came out with something decent in a few months.

>> No.3363942

>>3363867
>implying there's ANYTHING machines can't eventually do better than humans

Prepare your anus for the coming decade fuckboi.

>> No.3364272

>>3363885
CS student here too. Neural networks can't really rate art, but theres simple ways to solve that. As it looks like, he will only feed it with portraits, meaning that you can make a neural network identfy how much of the drawing resambles a human face and then rate it following a algorithm.

I'm sure that are other ways too, like by comparing a large number of pre-selected drawings already rated by a human to the input drawing and then apply the neural network on that.

Neural networks and deep learning are super powerful, they dont only generate images. Thats just a tiny demostration of what they can do.