Saved in:
Bibliographic Details
Main Authors: Walker, Paul, Gardner, James A. D., Ardelean, Andreea, Smith, William A. P., Egger, Bernhard
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2601.14079
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908776505802752
author Walker, Paul
Gardner, James A. D.
Ardelean, Andreea
Smith, William A. P.
Egger, Bernhard
author_facet Walker, Paul
Gardner, James A. D.
Ardelean, Andreea
Smith, William A. P.
Egger, Bernhard
contents Inverse rendering is an ill-posed problem, but priors like illumination priors, can simplify it. Existing work either disregards the spherical and rotation-equivariant nature of illumination environments or does not provide a well-behaved latent space. We propose a rotation-equivariant variational autoencoder that models natural illumination on the sphere without relying on 2D projections. To preserve the SO(2)-equivariance of environment maps, we use a novel Vector Neuron Vision Transformer (VN-ViT) as encoder and a rotation-equivariant conditional neural field as decoder. In the encoder, we reduce the equivariance from SO(3) to SO(2) using a novel SO(2)-equivariant fully connected layer, an extension of Vector Neurons. We show that our SO(2)-equivariant fully connected layer outperforms standard Vector Neurons when used in our SO(2)-equivariant model. Compared to previous methods, our variational autoencoder enables smoother interpolation in latent space and offers a more well-behaved latent space.
format Preprint
id arxiv_https___arxiv_org_abs_2601_14079
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle VENI: Variational Encoder for Natural Illumination
Walker, Paul
Gardner, James A. D.
Ardelean, Andreea
Smith, William A. P.
Egger, Bernhard
Computer Vision and Pattern Recognition
Inverse rendering is an ill-posed problem, but priors like illumination priors, can simplify it. Existing work either disregards the spherical and rotation-equivariant nature of illumination environments or does not provide a well-behaved latent space. We propose a rotation-equivariant variational autoencoder that models natural illumination on the sphere without relying on 2D projections. To preserve the SO(2)-equivariance of environment maps, we use a novel Vector Neuron Vision Transformer (VN-ViT) as encoder and a rotation-equivariant conditional neural field as decoder. In the encoder, we reduce the equivariance from SO(3) to SO(2) using a novel SO(2)-equivariant fully connected layer, an extension of Vector Neurons. We show that our SO(2)-equivariant fully connected layer outperforms standard Vector Neurons when used in our SO(2)-equivariant model. Compared to previous methods, our variational autoencoder enables smoother interpolation in latent space and offers a more well-behaved latent space.
title VENI: Variational Encoder for Natural Illumination
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2601.14079