Saved in:
Bibliographic Details
Main Authors: Zhang, Xiao, Gao, William, Jain, Seemandhar, Maire, Michael, Forsyth, David A., Bhattad, Anand
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2405.21074
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908303378874368
author Zhang, Xiao
Gao, William
Jain, Seemandhar
Maire, Michael
Forsyth, David A.
Bhattad, Anand
author_facet Zhang, Xiao
Gao, William
Jain, Seemandhar
Maire, Michael
Forsyth, David A.
Bhattad, Anand
contents Image relighting is the task of showing what a scene from a source image would look like if illuminated differently. Inverse graphics schemes recover an explicit representation of geometry and a set of chosen intrinsics, then relight with some form of renderer. However error control for inverse graphics is difficult, and inverse graphics methods can represent only the effects of the chosen intrinsics. This paper describes a relighting method that is entirely data-driven, where intrinsics and lighting are each represented as latent variables. Our approach produces SOTA relightings of real scenes, as measured by standard metrics. We show that albedo can be recovered from our latent intrinsics without using any example albedos, and that the albedos recovered are competitive with SOTA methods.
format Preprint
id arxiv_https___arxiv_org_abs_2405_21074
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Latent Intrinsics Emerge from Training to Relight
Zhang, Xiao
Gao, William
Jain, Seemandhar
Maire, Michael
Forsyth, David A.
Bhattad, Anand
Computer Vision and Pattern Recognition
Image relighting is the task of showing what a scene from a source image would look like if illuminated differently. Inverse graphics schemes recover an explicit representation of geometry and a set of chosen intrinsics, then relight with some form of renderer. However error control for inverse graphics is difficult, and inverse graphics methods can represent only the effects of the chosen intrinsics. This paper describes a relighting method that is entirely data-driven, where intrinsics and lighting are each represented as latent variables. Our approach produces SOTA relightings of real scenes, as measured by standard metrics. We show that albedo can be recovered from our latent intrinsics without using any example albedos, and that the albedos recovered are competitive with SOTA methods.
title Latent Intrinsics Emerge from Training to Relight
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2405.21074