1
$\begingroup$

I'm not sure this is the right place to ask for this reference but it's worth the shot since it is graphics related.

I'm looking for the very first paper describing what "Inverse Rendering" - By that I mean a machine learning system that given some renderer and a deep learning model and an image as input will provide you one or more maps describing how to render the same image.

I cannot manage to find anywhere a/the white paper the would describe how such system would work.

Can anyone provide a reference? Maybe not the very first paper but a white paper that might describe the fundamentals (I'd rather prefer papers to web articles).

Thank you

Improve this question
$\endgroup$

2 Answers 2

Reset to default
2
$\begingroup$

I only discovered there was name for it last Siggraph despite it being around, here is good introduction https://diff-render.org/ with lots of references to more specialized topics. Can somebody confirm this phd thesis the earliest work on it? https://www.graphics.cornell.edu/pubs/1998/Mar98.pdf

Improve this answer
$\endgroup$
1
$\begingroup$

I did only quick google search and found this PhD thesis - "Inverse rendering for computer graphics" by [Marschner1998] - http://graphics.stanford.edu/~srm/thesis/thesis.pdf

Also two published papers - Ramamoorthi et al "A Signal-Processing Framework for Inverse Rendering" http://cseweb.ucsd.edu/~ravir/papers/invrend/invrend.pdf

And Yu, Smith - "InverseRenderNet: Learning Single image inverse - https://openaccess.thecvf.com/content_CVPR_2019/papers/Yu_InverseRenderNet_Learning_Single_Image_Inverse_Rendering_CVPR_2019_paper.pdf

Based on my experience, just follow prior works in this area and you'll get to the first or first worth mentioning paper in this area. With every paper in the area of raytracing, you'll get reference to [Whitted1980]. Every BRDF will get you back to [CT1967] and every Spherical Harmonics work to [Green2003].

Based on what you wrote, you must be familiar with reading papers. Therefore, I would suggest to read "Introduction" and "Prior work" to find references to past works in this area of computer graphics.

Improve this answer
$\endgroup$
3
  • $\begingroup$ I found already the first two references you pointed out. But they're not deep learning based, I'm already aware of Whitted1980 and CT1967 but they have nothing to do with inverse rendering. Based on the thesis Inverse rendering it self has nothing to do with deep learning (I personally thought it was something enabled by Deep Learning, but I was wrong). I'm doing my research at the moment. The rendering network you pointed out is interesting, but it's dated 2019 and I have a feeling Inverse Rendering with Deep Learning started before that. But I can start with that I guess. $\endgroup$
    – user8469759
    Sep 11, 2020 at 12:26
  • 1
    $\begingroup$ I am aware that they're not related. It was just meant to be an example of "basic papers" in other areas of computer graphics. I'm not sure if you understood me correctly. What I meant was to use references at the end of papers to find related works. In Yu/Smith work from 2019, I found in reference section this paper - papers.neurips.cc/paper/… - Kulkarni et al 2015 - Deep Convolutional Inverse Graphics Network. $\endgroup$
    – DirectX_Programmer
    Sep 11, 2020 at 15:58
  • 1
    $\begingroup$ I understood what you meant, essentially start from whatever paper I find and navigate backwards through the references until I find the first paper. $\endgroup$
    – user8469759
    Sep 11, 2020 at 16:04

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.