Dumb layman question

[panhartstuff]

I want to ask a bit of a dumb question: why isn't this a full implementation of what we see in the tech demo video? Especially when Nvidia have released the whole code? Why does further development still needed to be done?

[noyoshi]

They did not release the whole code - they just released some code to test and evaluate a few pretrained models. In the demo, they are using the Flickr - trained model, which they did not release, and they are running it on a very very expensive computer with many GPUs, to achieve near real time image synthesis. The code they released is very much research - oriented, and is not at all written to be used in a user-facing application.

However, I'll add that getting to an app that functions like their demo is a function of time and money, as technically everything they used can be acquired now. There is some talk about training an open-sourced version of the Flickr model, as the dataset is open source, and if someone has a mammoth computer to run the models on, we could in theory make something like what was in the demo. I think this might be a waste of time though, as someone from NVIDIA said they would be making and releasing their own demo app in the near future.

[energy888666]

Now does NVIDIA publish the whole code ?