Differentiable Rendering with Reparameterized Volume Sampling

Nikita Morozov, Denis Rakitin, Oleg Desheulin, Dmitry P Vetrov, Kirill Struminsky
Proceedings of The 27th International Conference on Artificial Intelligence and Statistics, PMLR 238:4852-4860, 2024.

Abstract

In view synthesis, a neural radiance field approximates underlying density and radiance fields based on a sparse set of scene pictures. To generate a pixel of a novel view, it marches a ray through the pixel and computes a weighted sum of radiance emitted from a dense set of ray points. This rendering algorithm is fully differentiable and facilitates gradient-based optimization of the fields. However, in practice, only a tiny opaque portion of the ray contributes most of the radiance to the sum. We propose a simple end-to-end differentiable sampling algorithm based on inverse transform sampling. It generates samples according to the probability distribution induced by the density field and picks non-transparent points on the ray. We utilize the algorithm in two ways. First, we propose a novel rendering approach based on Monte Carlo estimates. This approach allows for evaluating and optimizing a neural radiance field with just a few radiance field calls per ray. Second, we use the sampling algorithm to modify the hierarchical scheme proposed in the original NeRF work. We show that our modification improves reconstruction quality of hierarchical models, at the same time simplifying the training procedure by removing the need for auxiliary proposal network losses.

Cite this Paper

BibTeX
@InProceedings{pmlr-v238-morozov24a, title = { Differentiable Rendering with Reparameterized Volume Sampling }, author = {Morozov, Nikita and Rakitin, Denis and Desheulin, Oleg and P Vetrov, Dmitry and Struminsky, Kirill}, booktitle = {Proceedings of The 27th International Conference on Artificial Intelligence and Statistics}, pages = {4852--4860}, year = {2024}, editor = {Dasgupta, Sanjoy and Mandt, Stephan and Li, Yingzhen}, volume = {238}, series = {Proceedings of Machine Learning Research}, month = {02--04 May}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v238/morozov24a/morozov24a.pdf}, url = {https://proceedings.mlr.press/v238/morozov24a.html}, abstract = { In view synthesis, a neural radiance field approximates underlying density and radiance fields based on a sparse set of scene pictures. To generate a pixel of a novel view, it marches a ray through the pixel and computes a weighted sum of radiance emitted from a dense set of ray points. This rendering algorithm is fully differentiable and facilitates gradient-based optimization of the fields. However, in practice, only a tiny opaque portion of the ray contributes most of the radiance to the sum. We propose a simple end-to-end differentiable sampling algorithm based on inverse transform sampling. It generates samples according to the probability distribution induced by the density field and picks non-transparent points on the ray. We utilize the algorithm in two ways. First, we propose a novel rendering approach based on Monte Carlo estimates. This approach allows for evaluating and optimizing a neural radiance field with just a few radiance field calls per ray. Second, we use the sampling algorithm to modify the hierarchical scheme proposed in the original NeRF work. We show that our modification improves reconstruction quality of hierarchical models, at the same time simplifying the training procedure by removing the need for auxiliary proposal network losses. } }
Endnote
%0 Conference Paper %T Differentiable Rendering with Reparameterized Volume Sampling %A Nikita Morozov %A Denis Rakitin %A Oleg Desheulin %A Dmitry P Vetrov %A Kirill Struminsky %B Proceedings of The 27th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2024 %E Sanjoy Dasgupta %E Stephan Mandt %E Yingzhen Li %F pmlr-v238-morozov24a %I PMLR %P 4852--4860 %U https://proceedings.mlr.press/v238/morozov24a.html %V 238 %X In view synthesis, a neural radiance field approximates underlying density and radiance fields based on a sparse set of scene pictures. To generate a pixel of a novel view, it marches a ray through the pixel and computes a weighted sum of radiance emitted from a dense set of ray points. This rendering algorithm is fully differentiable and facilitates gradient-based optimization of the fields. However, in practice, only a tiny opaque portion of the ray contributes most of the radiance to the sum. We propose a simple end-to-end differentiable sampling algorithm based on inverse transform sampling. It generates samples according to the probability distribution induced by the density field and picks non-transparent points on the ray. We utilize the algorithm in two ways. First, we propose a novel rendering approach based on Monte Carlo estimates. This approach allows for evaluating and optimizing a neural radiance field with just a few radiance field calls per ray. Second, we use the sampling algorithm to modify the hierarchical scheme proposed in the original NeRF work. We show that our modification improves reconstruction quality of hierarchical models, at the same time simplifying the training procedure by removing the need for auxiliary proposal network losses.
APA
Morozov, N., Rakitin, D., Desheulin, O., P Vetrov, D. & Struminsky, K.. (2024). Differentiable Rendering with Reparameterized Volume Sampling . Proceedings of The 27th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 238:4852-4860 Available from https://proceedings.mlr.press/v238/morozov24a.html.

Related Material