Training @NvidiaAI's #StyleGAN on Google Earth sattelite imagery, will take a few more days of training to get good samples, but already looks promising! Unfortunately don't have enough compute budget to run the model at full 1024 resolution...

5:15 PM · Feb 16, 2019

9
46
7
233
GIF
Real samples vs Generated samples after 24 hours of training. Gonna wait another day to generate latent space interpolations, let's not spoil all the fun at once ;-)
5
8
0
33
Final generator samples after training #StyleGAN on satellite imagery for 3 days using 2xV100 GPUs at 256x256 resolution. Absolutely love the variety & esthetics of the results!
2
10
0
32
Next, here are some mandatory slerp-interpolations in latent 'w' space, using some code from @phillip_isola gist.github.com/phillipi/d29… GIF loops between 3 latent midpoints and finally crosses the mean of the distribution since loopable GIFs are dope 😋
1
1
0
8
GIF
And finally, @gwern, here are some rather artsy-looking high-resolution samples generated with ESRGAN. The super-resolution GAN adds some funny artefacts which makes the result kinda look like oil paintings. (didn't use these for the interpolations since they introduced jitter)
1
0
0
4
Can't wait for your travels through latent space!
0
0
0
4
You’re unable to view this Tweet because this account owner limits who can view their Tweets.
Hmm sweet, didn't think of that yet! You think the pre-trained model will generalize well to satellite imagery? Also for the resolution, I haven't actually checked whether I can initialize a 1024-model with my current 256-one and then continue training.. Not sure if that works 🤔
1
0
0
0
It’ll be interesting to see a StyleGAN version of @golan’s “Terrapattern” project based on generated satellite imagery!
After two years, terrapattern.com, our ML-powered prototype for reverse image search in satellite imagery, will shut down in May. Give it a spin while you can, or check out the demo video, invidio.us/watch?v=AErDoe-5…
0
0
1
11
Could you give a link to the dataset?
0
0
0
0
What budget do you need and what GPU on AWS ?
1
0
0
2
I'm using 2x V100 GPUs with 12 cpu-cores and 100Gb SSD memory, a little over 100$/day of training. Training the full 1024-resolution model costs roughly 2k$ when using cloud compute..
1
0
1
0