Despite recent advances in Monte Carlo path tracing at interactive rates, denoised image
sequences generated with few samples per‐pixel often yield temporally unstable results and
loss of high‐frequency details. We present a novel adaptive rendering method that
increases temporal stability and image fidelity of low sample count path tracing by
distributing samples via spatio‐temporal joint optimization of sampling and denoising.
Adding temporal optimization to the sample predictor enables it to learn spatio‐temporal …

We have experimented with a smaller versions of the networks detailed in the paper,
targeted at real-time applications with a runtime of about 13 ms per frame at 1920× 1080. In
our experience, image quality scales gracefully with network size. It is hard to claim an
optimal design, and it rather depends on run-time constraints or a subjective notion of how
much quality can be sacrificed for performance. In particular, we note that the sampler
network can be aggressively downscaled without noticeable impact on image quality. We …