DLSS: Image Reconstruction for Real-time Rendering with Deep Learning

First presented as part of GTC Digital, NVIDIA Senior Research Scientist Edward Liu discussed the latest research progress on Deep Learning Super Sampling (DLSS), which uses deep learning and NVIDIA Tensor Cores to reconstruct super sampled frames in real-time.

“Over the last year, we have improved every aspect of DLSS – from the training methodology, loss function, the dataset, how the algorithm interacts with a game engine, and ultimately redesigning the entire architecture,” says Edward Liu.

He further demonstrates why scaling and image reconstruction for real-time rendering is an extremely challenging problem, and examines the cause of common artifacts produced by traditional up-sampling techniques.

Watch the first five minutes of the presentation here:

(note: you must register for the NVIDIA Developer Program to view the full video).

To learn more about DLSS 2.0 or to request access to DLSS 2.0 in Unreal Engine, visit our page here