“Learning Light Transport the Reinforced Way” by Dahm and Keller
Conference:
Type:
Entry Number: 73
Title:
- Learning Light Transport the Reinforced Way
Presenter(s)/Author(s):
Abstract:
We introduce a rendering algorithm that is as simple as a path tracer but dramatically improves light transport simulation and even outperforms the Metropolis light transport algorithm. The underlying method of importance sampling learns where radiance is coming from and in fact coincides with reinforcement learning. The cost for the improvement is a data structure similar to irradiance volumes as used in realtime games.
References:
K. Dahm and A. Keller. 2017. Learning Light Transport the Reinforced Way. CoRR abs/1701.07403 (2017). http://arxiv.org/abs/1701.07403Google Scholar
G. Greger, P. Shirley, P. Hubbard, and D. Greenberg. 1998. The Irradiance Volume. IEEE Computer Graphics and Applications 18, 2 (1998), 32–43. Google ScholarDigital Library
C. Kelemen, L. Szirmay-Kalos, G. Antal, and F. Csonka. 2002. A Simple and Robust Mutation Strategy for the Metropolis Light Transport Algorithm. Computer Graphics Forum 21, 3 (2002), 531–540. Google ScholarCross Ref
E. Lafortune and Y. Willems. 1995. A 5D Tree to Reduce the Variance of Monte Carlo Ray Tracing. In Rendering Techniques 1995 (Proc. 6th Eurographics Workshop on Rendering), P. Hanrahan and W. Purgathofer (Eds.). Springer, 11–20.Google Scholar
M. Pharr, W. Jacob, and G. Humphreys. 2016. Physically Based Rendering – From Theory to Implementation. Morgan Kaufmann, Third Edition.Google Scholar
E. Veach. 1997. Robust Monte Carlo Methods for Light Transport Simulation. Ph.D. Dissertation. Stanford University.Google Scholar
E. Veach and L. Guibas. 1997. Metropolis Light Transport. In Proc. SIGGRAPH 1997 (Annual Conference Series), Turner Whitted (Ed.). ACM SIGGRAPH, Addison Wesley, 65–76. Google ScholarDigital Library
J. Vorba, O. Karlík, M. Šik, T. Ritschel, and J. Křivánek. 2014. On-line Learning of Parametric Mixture Models for Light Transport Simulation. ACM Transactions on Graphics (Proceedings of SIGGRAPH 2014) 33, 4 (2014).Google Scholar
C. Watkins and P. Dayan. 1992. Q-Learning. Machine learning 8, 3 (1992), 279–292.
Keyword(s):
- _rendering
- Path Tracing
- reinforcement learning
- integral equations
- Monte Carlo
- quasi-Monte Carlo methods
Acknowledgements:
The authors would like to thank Jaroslav Kriv ˇ anek and Tero Karras ́ for profound discussions. Scene courtesy (cc) 2013 Miika Ai”ala, Samuli Laine, and Jaakko Lehtinen, see https://mediatech.aalto.#/ publications/graphics/GMLT/.
