Papers of the day   All papers

DeepLight: Learning Illumination for Unconstrained Mobile Mixed Reality


May 20 2019 Mean Squared Error

DeepLight (SIGGRAPH 2019) presents a system to infer plausible omni-directional HDR lighting at interactive rates, using LDR captures from narrow FoV mobile phone cameras. DeepLight is the basis of Google's upgraded lighting estimation feature in ARCore 2
2 replies, 151 likes

May 05 2019 Paul Debevec

Some of our AR lighting technology will debut at Google I/O next week; research paper on the technique to appear at CVPR 2019
1 replies, 65 likes

May 20 2019 Dimitri Diakopoulos

Hallucinating a full HDR IBL from a single image is a neat trick. This thread of research is equally (or more) important for AR headsets.
0 replies, 32 likes

Apr 04 2019 Simon Green

DeepLight. Basically, learn to predict the lighting environment by sticking diffuse and mirrored spheres in front of the camera!
1 replies, 24 likes

Apr 11 2019 Paul Debevec

Thank you @jflalondeqc and for your impressive and groundbreaking work with learning-based lighting estimation as well!
0 replies, 5 likes

Apr 10 2019 Jean-Francois Lalonde

Cool work from @debfx's team coming up at @cvpr2019: DeepLight: Learning Illumination for Unconstrained Mobile Mixed Reality. Great way to gather lots of illumination data!
0 replies, 4 likes

Jun 20 2019 Alex Ma

Today 10-11am we are going to do a live demo at the @cvpr2019 Google booth for the machine learning based AR lighting estimation Looking forward to meeting you at the exhibition!
0 replies, 4 likes

Aug 19 2019 Tobias A. Franke

"DeepLight: Learning Illumination for Unconstrained Mobile Mixed Reality" Interesting way to generate environment maps for #AugmentedReality reflection probes from the camera's limited input image.
0 replies, 1 likes