Looks absolutely gorgeous!
Very interesting post, thanks for sharing! Next step would be to procedurally create the indoors! I understand your point about a semi-procedural approach. I often fall into the go-full-procedural vortex and can't get things done. :D Anyway, one question I don't quite get why "fill the interior" on the vdb node has to be checked?
it is really frustrating that half the article was posted and i cannot view the end of this article, every possible link on the page refers to the same url and that url is password protected. creation of a gametextures account will not get you access to the end of this page so don't hand them paypal info like I did just go blunder off through a youtube tutorial rather than this good-old-boy referral page
NVIDIA has presented a new deep learning-based approach that can allow users to use photos that were originally taken in low light and automatically remove the noise and artifacts. The project is said to be developed by researchers from NVIDIA, Aalto University, and MIT.
This method might not seem unique, but the key thing key is that it only takes two input images with the noise or grain to get great results. This AI doesn’t even need to study a noise-free image to remove artifacts, noise, grain, and automatically enhance photos.
“It is possible to learn to restore signals without ever observing clean ones, at performance sometimes exceeding training using clean exemplars,” the researchers pointed out.“[The neural network] is on par with state-of-the-art methods that make use of clean examples — using precisely the same training methodology, and often without appreciable drawbacks in training time or performance.”
The team validated the neural network on three different datasets to check if the system does the work. The method is said to be capable of enhancing MRI images, so medical imaging has a great future.
“There are several real-world situations where obtaining clean training data is difficult: low-light photography (e.g., astronomical imaging), physically-based rendering, and magnetic resonance imaging,” the team stated. “Our proof-of-concept demonstrations point the way to significant potential benefits in these applications by removing the need for potentially strenuous collection of clean data. Of course, there is no free lunch – we cannot learn to pick up features that are not there in the input data – but this applies equally to training with clean targets.”
You can learn more about the paper here.