Photon Field Networks for Dynamic Real-Time Volumetric Global Illumination

David Bauer, Qi Wu, Kwan-Liu Ma

Room: 106

2023-10-26T00:33:00ZGMT-0600Change your timezone on the schedule page
Exemplar figure, described by caption below
Photon Field Networks are implicit neural represenations trained on photon caches traced in volume datasets. They can represent indirect radiance parameterized for sample position and view direction. Moreover, they can be trained on multiple caches simultaneously to learn non-isotropic scattering effects. In this paper, we introduce the concept of Photon Fields, show how to efficiently train them, and evaluate them in a proof-of-concept path tracing application. Our results show that Photon Fields can faithfully represent the photon caches and create approximate global illumination effects several times faster than a comparable path tracer.
Fast forward
Full Video

Volume data, volume rendering, volume visualization, deep learning, global illumination, neural rendering, path tracing


Volume data is commonly found in many scientific disciplines, like medicine, physics, and biology. Experts rely on robust scientific visualization techniques to extract valuable insights from the data. Recent years have shown path tracing to be the preferred approach for volumetric rendering, given its high levels of realism. However, real-time volumetric path tracing often suffers from stochastic noise and long convergence times, limiting interactive exploration. In this paper, we present a novel method to enable real-time global illumination for volume data visualization. We develop Photon Field Networks - a phase-function-aware, multi-light neural representation of indirect volumetric global illumination. The fields are trained on multi-phase photon caches that we compute a priori. Training can be done within seconds, after which the fields can be used in various rendering tasks. To showcase their potential, we develop a custom neural path tracer, with which our photon fields achieve interactive framerates even on large datasets. We conduct in-depth evaluations of the method's performance, including visual quality, stochastic noise, inference and rendering speeds, and accuracy regarding illumination and phase function awareness. Results are compared to ray marching, path tracing and photon mapping. Our findings show that Photon Field Networks can faithfully represent indirect global illumination within the boundaries of the trained phase spectrum while exhibiting less stochastic noise and rendering at a significantly faster rate than traditional methods.