Show HN: GAN Playground – Play with Generative Adversarial Neural Nets In-Browser (opens in new tab)

(reiinakano.github.io)

97 pointsreiinakano8y ago14 comments

14 comments

13 comments · 3 top-level

SonOfLilit8y ago· 9 in thread

I gave it a few minutes and got nothing.

Maybe try for something even simpler than MNIST, so we can get the great immediate feedback effect of http://playground.tensorflow.org, which I consider the important aspect for learnability?

sp3328y ago

There's just no point getting fancy for MNIST. Using nothing but fully-connected layers and rmsprop, I got good results in less than one minute. https://imgur.com/a/7xzmQ

Hidden units for discriminator layers - 100, 40, 2.

Generator layers - 40, 100, 768, then "reshape" into 28x28x1.

reiinakanoOP8y ago

Unlike regular classification, it's hard to quantify the performance of a GAN by just looking at the loss curves. You really have to make a subjective evaluation by looking at the generated images.

By playing around with it, the results I got from fully connected were nowhere near as good as the results I got from convolutional.

EvgeniyZh8y ago

what you're showing is basically perfect discriminator with awful generator, isn't it?

sp3328y ago

Yeah ok it's pretty misleading. And if you let it run longer the generator starts getting a little better. It took me a long time to figure out the UI.

1 more reply

reiinakanoOP8y ago

Default parameters on MNIST should give good results in 5 minutes give or take - depending on your hardware. Can you show what you got?

EvgeniyZh8y ago

That's what I got after 10 minutes on GTX1080 - https://imgur.com/a/4Ba9u

reiinakanoOP8y ago

Looks about right. The default setting is as basic as you can get. No transpose convolutions, batch norm, and other tricks for better GANs.

DiThi8y ago

Did you click "train"? Otherwise it's just doing inference from random data.

SonOfLilit8y ago

Thanks. Doh.

In Israel we have a classic song that goes "if you're cooking spaghetti and the water doesn't boil, how about turning the stove on, because that's what everybody does"

gallerdude8y ago· 1 in thread

Small nitpick, but I feel like it should start training by default. Took me a good minute to realize why no progress was being made.

zardo8y ago

It does for me in Firefox, but I have the opposite opinion. It shouldn't start running compute intensive operations and downloading datasets untill I say go.

rahimnathwani8y ago

Source and README: https://github.com/reiinakano/gan-playground

(at the time of writing, the HN post points to the demo)

j / k navigate · click thread line to collapse

14 comments

13 comments · 3 top-level

SonOfLilit8y ago· 9 in thread

I gave it a few minutes and got nothing.

Maybe try for something even simpler than MNIST, so we can get the great immediate feedback effect of http://playground.tensorflow.org, which I consider the important aspect for learnability?

sp3328y ago

There's just no point getting fancy for MNIST. Using nothing but fully-connected layers and rmsprop, I got good results in less than one minute. https://imgur.com/a/7xzmQ

Hidden units for discriminator layers - 100, 40, 2.

Generator layers - 40, 100, 768, then "reshape" into 28x28x1.

reiinakanoOP8y ago

Unlike regular classification, it's hard to quantify the performance of a GAN by just looking at the loss curves. You really have to make a subjective evaluation by looking at the generated images.

By playing around with it, the results I got from fully connected were nowhere near as good as the results I got from convolutional.

EvgeniyZh8y ago

what you're showing is basically perfect discriminator with awful generator, isn't it?

sp3328y ago

Yeah ok it's pretty misleading. And if you let it run longer the generator starts getting a little better. It took me a long time to figure out the UI.

1 more reply

reiinakanoOP8y ago

Default parameters on MNIST should give good results in 5 minutes give or take - depending on your hardware. Can you show what you got?

EvgeniyZh8y ago

That's what I got after 10 minutes on GTX1080 - https://imgur.com/a/4Ba9u

reiinakanoOP8y ago

Looks about right. The default setting is as basic as you can get. No transpose convolutions, batch norm, and other tricks for better GANs.

DiThi8y ago

Did you click "train"? Otherwise it's just doing inference from random data.

SonOfLilit8y ago

Thanks. Doh.

In Israel we have a classic song that goes "if you're cooking spaghetti and the water doesn't boil, how about turning the stove on, because that's what everybody does"

gallerdude8y ago· 1 in thread

Small nitpick, but I feel like it should start training by default. Took me a good minute to realize why no progress was being made.

zardo8y ago

It does for me in Firefox, but I have the opposite opinion. It shouldn't start running compute intensive operations and downloading datasets untill I say go.

rahimnathwani8y ago

Source and README: https://github.com/reiinakano/gan-playground

(at the time of writing, the HN post points to the demo)

j / k navigate · click thread line to collapse