A neat fact about this effort: the model was trained as a collaboration with a group of strangers who found each other on a Github issue page (https://github.com/NVlabs/SPADE/issues/17 …). We can match the big players in this field by pooling our resources together 


-
-
Prikaži ovu nitHvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
This is so awesome!!! Where'd you get all the labeled data, I thought that to be the main advantage of NVIDIA ( aside from a bunch of GPUs). That's really cool
-
we basically retraced the steps they wrote in the paper. we scraped a bunch of images from flickr and then labeled them with the cocostuffs-164k pretrained model fromhttps://github.com/kazuto1011/deeplab-pytorch/ …
- Još 1 odgovor
Novi razgovor -
-
-
Great job! Any estimate when we can try GauGAN in
@RunwayML? Also are you publishing the GUI you show in your video? -
We should have it up this week. There will be a similar gui in runway
- Još 1 odgovor
Novi razgovor -
-
-
Hi Gene, really nice work! I tried to play with your pretrained model with the original SPADE code but always run into a weird CUDA error. Would you mind to share a bit more details about test setting that you use to run your pretrained model? Thanks
-
maybe you're not setting flags? make sure it matches train settings. i'm not sure which model you mean, but if landscapes, then you need --load_size 1024 --crop_size 1024 --aspect_ratio 1.0 --no_instance --label_nc 182 --preprocess_mode scale_width --dataset_mode custom
- Još 1 odgovor
Novi razgovor -
-
-
kinda Bob Ross-esque...
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
We need this in
@Procreate now
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
looking forward
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.