S3R-Net: A Single-Stage Approach to Self-Supervised Shadow Removal

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Kubiak, Nikolina, Mustafa, Armin, Phillipson, Graeme, Jolly, Stephen, Hadfield, Simon
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866910414565015552
author Kubiak, Nikolina
Mustafa, Armin
Phillipson, Graeme
Jolly, Stephen
Hadfield, Simon
author_facet Kubiak, Nikolina
Mustafa, Armin
Phillipson, Graeme
Jolly, Stephen
Hadfield, Simon
contents In this paper we present S3R-Net, the Self-Supervised Shadow Removal Network. The two-branch WGAN model achieves self-supervision relying on the unify-and-adaptphenomenon - it unifies the style of the output data and infers its characteristics from a database of unaligned shadow-free reference images. This approach stands in contrast to the large body of supervised frameworks. S3R-Net also differentiates itself from the few existing self-supervised models operating in a cycle-consistent manner, as it is a non-cyclic, unidirectional solution. The proposed framework achieves comparable numerical scores to recent selfsupervised shadow removal models while exhibiting superior qualitative performance and keeping the computational cost low.
format Preprint
id arxiv_https___arxiv_org_abs_2404_12103
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle S3R-Net: A Single-Stage Approach to Self-Supervised Shadow Removal
Kubiak, Nikolina
Mustafa, Armin
Phillipson, Graeme
Jolly, Stephen
Hadfield, Simon
Computer Vision and Pattern Recognition
Graphics
In this paper we present S3R-Net, the Self-Supervised Shadow Removal Network. The two-branch WGAN model achieves self-supervision relying on the unify-and-adaptphenomenon - it unifies the style of the output data and infers its characteristics from a database of unaligned shadow-free reference images. This approach stands in contrast to the large body of supervised frameworks. S3R-Net also differentiates itself from the few existing self-supervised models operating in a cycle-consistent manner, as it is a non-cyclic, unidirectional solution. The proposed framework achieves comparable numerical scores to recent selfsupervised shadow removal models while exhibiting superior qualitative performance and keeping the computational cost low.
title S3R-Net: A Single-Stage Approach to Self-Supervised Shadow Removal
topic Computer Vision and Pattern Recognition
Graphics
url https://arxiv.org/abs/2404.12103