Retraining Optix AI denoiser with custom dataset

I want to retrain the network of Optix denoiser with custom dataset.

When I saw a programming guide, I found that retraining is possible and Optix provides custom training data buffer.
As Optix didn’t support an interface for training, I referred to the course “Rendered Image Denoising using Autoencoders”.
Although the course provided training code, It was built on Qwiklabs, a cloud platform.
So I can’t put my custom inputs and get binary training files.

Is there way to get training code directly?
Or, how can I leverage Qwiklabs course for retraining network with custom dataset?

Thanks.

I’m interested to train my test data too.
thanks

Hello, I am also interested to train the model with my own data. I also do know if the nvidia solution creates a images with RGB, Albedo and Normal channels or I have to do it by myself. The course do not provide the roadmap to convert the images to array numpy. Thanks

OptiX 5.0.1 had a folder “denoiser_training” in folder “tool”
[url]https://developer.nvidia.com/designworks/optix/downloads/legacy[/url]
Maybe that helps. However, I don’t know whether that code still works with OptiX 6.0.0

Denoiser: [url]https://developer.nvidia.com/optix-denoiser[/url]
Recurrent Denoising Autoencoder paper: [url]http://research.nvidia.com/publication/interactive-reconstruction-monte-carlo-image-sequences-using-recurrent-denoising[/url]

Thanks for answers.
I will check the reference.

Thanks.