NeRF-VAE: A Geometry Aware 3D Scene Generative Model

Adam R Kosiorek, Heiko Strathmann, Daniel Zoran, Pol Moreno, Rosalia Schneider, Sona Mokra, Danilo Jimenez Rezende
Proceedings of the 38th International Conference on Machine Learning, PMLR 139:5742-5752, 2021.

Abstract

We propose NeRF-VAE, a 3D scene generative model that incorporates geometric structure via Neural Radiance Fields (NeRF) and differentiable volume rendering. In contrast to NeRF, our model takes into account shared structure across scenes, and is able to infer the structure of a novel scene—without the need to re-train—using amortized inference. NeRF-VAE’s explicit 3D rendering process further contrasts previous generative models with convolution-based rendering which lacks geometric structure. Our model is a VAE that learns a distribution over radiance fields by conditioning them on a latent scene representation. We show that, once trained, NeRF-VAE is able to infer and render geometrically-consistent scenes from previously unseen 3D environments of synthetic scenes using very few input images. We further demonstrate that NeRF-VAE generalizes well to out-of-distribution cameras, while convolutional models do not. Finally, we introduce and study an attention-based conditioning mechanism of NeRF-VAE’s decoder, which improves model performance.

Cite this Paper

BibTeX
@InProceedings{pmlr-v139-kosiorek21a, title = {NeRF-VAE: A Geometry Aware 3D Scene Generative Model}, author = {Kosiorek, Adam R and Strathmann, Heiko and Zoran, Daniel and Moreno, Pol and Schneider, Rosalia and Mokra, Sona and Rezende, Danilo Jimenez}, booktitle = {Proceedings of the 38th International Conference on Machine Learning}, pages = {5742--5752}, year = {2021}, editor = {Meila, Marina and Zhang, Tong}, volume = {139}, series = {Proceedings of Machine Learning Research}, month = {18--24 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v139/kosiorek21a/kosiorek21a.pdf}, url = {https://proceedings.mlr.press/v139/kosiorek21a.html}, abstract = {We propose NeRF-VAE, a 3D scene generative model that incorporates geometric structure via Neural Radiance Fields (NeRF) and differentiable volume rendering. In contrast to NeRF, our model takes into account shared structure across scenes, and is able to infer the structure of a novel scene—without the need to re-train—using amortized inference. NeRF-VAE’s explicit 3D rendering process further contrasts previous generative models with convolution-based rendering which lacks geometric structure. Our model is a VAE that learns a distribution over radiance fields by conditioning them on a latent scene representation. We show that, once trained, NeRF-VAE is able to infer and render geometrically-consistent scenes from previously unseen 3D environments of synthetic scenes using very few input images. We further demonstrate that NeRF-VAE generalizes well to out-of-distribution cameras, while convolutional models do not. Finally, we introduce and study an attention-based conditioning mechanism of NeRF-VAE’s decoder, which improves model performance.} }
Endnote
%0 Conference Paper %T NeRF-VAE: A Geometry Aware 3D Scene Generative Model %A Adam R Kosiorek %A Heiko Strathmann %A Daniel Zoran %A Pol Moreno %A Rosalia Schneider %A Sona Mokra %A Danilo Jimenez Rezende %B Proceedings of the 38th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2021 %E Marina Meila %E Tong Zhang %F pmlr-v139-kosiorek21a %I PMLR %P 5742--5752 %U https://proceedings.mlr.press/v139/kosiorek21a.html %V 139 %X We propose NeRF-VAE, a 3D scene generative model that incorporates geometric structure via Neural Radiance Fields (NeRF) and differentiable volume rendering. In contrast to NeRF, our model takes into account shared structure across scenes, and is able to infer the structure of a novel scene—without the need to re-train—using amortized inference. NeRF-VAE’s explicit 3D rendering process further contrasts previous generative models with convolution-based rendering which lacks geometric structure. Our model is a VAE that learns a distribution over radiance fields by conditioning them on a latent scene representation. We show that, once trained, NeRF-VAE is able to infer and render geometrically-consistent scenes from previously unseen 3D environments of synthetic scenes using very few input images. We further demonstrate that NeRF-VAE generalizes well to out-of-distribution cameras, while convolutional models do not. Finally, we introduce and study an attention-based conditioning mechanism of NeRF-VAE’s decoder, which improves model performance.
APA
Kosiorek, A.R., Strathmann, H., Zoran, D., Moreno, P., Schneider, R., Mokra, S. & Rezende, D.J.. (2021). NeRF-VAE: A Geometry Aware 3D Scene Generative Model. Proceedings of the 38th International Conference on Machine Learning, in Proceedings of Machine Learning Research 139:5742-5752 Available from https://proceedings.mlr.press/v139/kosiorek21a.html.

Related Material