Neural model improves light reflection sampling for realistic rendering
ART-MPG: Autoregressive R/T Manifold Path Guiding
Machine Learning
Summary
Rendering realistic images on computers involves simulating how light bounces and reflects in a scene, which can be very complex. The paper presents a new approach called ART-MPG that uses a compact neural model to better predict how light paths behave, especially where shiny surfaces cause tricky reflections. This method reduces the number of repeated attempts needed to correctly sample these light paths, making rendering both more accurate and efficient. The authors tested their method on many scenes and found it consistently lowers errors compared to existing techniques.
What this means in practice
- •For graphics engineers: Enhance photorealistic rendering engines by reducing noise and improving sampling efficiency of complex light paths involving reflections and transmissions.
- •For visual effects teams: Produce more accurate renderings of scenes with shiny materials and complex light interactions using improved path guiding for faster iteration.
Authors
Zehao Zhao, Yanyang Yang, Xiaole Zhang, Xiyu Deng
Abstract
Specular chains concentrate important transport in small regions of path space and the structure depends on chain topology. We introduce ART-MPG, a compact neural model that samples chain length, an autoregressive reflection/transmission (R/T) string, and a direction from a topology-conditioned four-component von Mises-Fisher (vMF) mixture. We extend the repeated trial weighting in Manifold Path Guiding (MPG) by conditioning retry sampling on both chain length and the sampled R/T string, accounting for its known mass outside the stochastic re-hit trials. For a fixed proposal and target chain, this preserves the ideal expectation while reducing the expected retry count and reciprocal-estimator variance compared with retries conditioned on chain length alone. We distinguish the formal unbiasedness result from the capped and clamped renderer used in experiments. We evaluate fourteen scenes across a range of samples per pixel (SPP) against existing online and offline rendering algorithms. Our method consistently produces lower mean squared error at equal SPP than the online and offline baselines, with lower standard deviation across runs.