Computer Science > Graphics
[Submitted on 26 Oct 2017 (v1), last revised 22 May 2018 (this version, v2)]
Title:Deep Illumination: Approximating Dynamic Global Illumination with Generative Adversarial Network
View PDFAbstract:We present Deep Illumination, a novel machine learning technique for approximating global illumination (GI) in real-time applications using a Conditional Generative Adversarial Network. Our primary focus is on generating indirect illumination and soft shadows with offline rendering quality at interactive rates. Inspired from recent advancement in image-to-image translation problems using deep generative convolutional networks, we introduce a variant of this network that learns a mapping from Gbuffers (depth map, normal map, and diffuse map) and direct illumination to any global illumination solution. Our primary contribution is showing that a generative model can be used to learn a density estimation from screen space buffers to an advanced illumination model for a 3D environment. Once trained, our network can approximate global illumination for scene configurations it has never encountered before within the environment it was trained on. We evaluate Deep Illumination through a comparison with both a state of the art real-time GI technique (VXGI) and an offline rendering GI technique (path tracing). We show that our method produces effective GI approximations and is also computationally cheaper than existing GI techniques. Our technique has the potential to replace existing precomputed and screen-space techniques for producing global illumination effects in dynamic scenes with physically-based rendering quality.
Submission history
From: Manu Mathew Thomas [view email][v1] Thu, 26 Oct 2017 17:56:00 UTC (6,273 KB)
[v2] Tue, 22 May 2018 23:16:53 UTC (6,273 KB)
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
Connected Papers (What is Connected Papers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.