Jump to content
art3mis

SIGGRAPH - Neural networks and the future of ray tracing

Recommended Posts

marty    574

I call bogus on this- it's a GPu denoiser. There are no denoisers that take minutes per frame:

Quote

Existing algorithms for high-quality denoising consume seconds to minutes per frame, which makes them impractical for interactive applications.

 

Share this post


Link to post
Share on other sites
acey195    109
5 hours ago, marty said:

I call bogus on this- it's a GPu denoiser. There are no denoisers that take minutes per frame:

 

Well it depends on what resolution you have, and what amount of data loss you allow (blurring)

I think predicting the location of noise and having data sets for image completion (to replace the noise with) is certainly not a bad idea...

There was also a siggraph demo of real-time fluids based on a similar principle

Share this post


Link to post
Share on other sites
marty    574

redshift devs looked at it and say that it helps for the last 5% maybe. It's just not very good at the moment.

adding that Neat Denoiser works at approx 4 fps for 2K images IIRC, 1080ti 12 cores@3.33, so a render at pixel samples of 5x5 would equal 25 times slower to denoise, which is 6.25 secs per frame.

Edited by marty
  • Like 1

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×