Logo do repositório
 
Publicação

SCENE-Net

2025-12Artigo científico Acesso aberto
Ver registo simples
dc.contributor.authorLavado, Diogo
dc.contributor.authorMicheletti, Alessandra
dc.contributor.authorBocchi, Giovanni
dc.contributor.authorFrosini, Patrizio
dc.contributor.authorSoares, Cláudia
dc.contributor.institutionFaculdade de Ciências e Tecnologia (FCT)
dc.contributor.pblAcademic Press Inc.
dc.date.accessioned2026-01-26T15:13:01Z
dc.date.available2026-01-26T15:13:01Z
dc.date.issued2025-12
dc.descriptionPublisher Copyright: © 2025 The Author(s).
dc.description.abstractThis paper introduces SCENE-Net, a novel low-resource, white-box model that serves as a compelling proof-of-concept for 3D point cloud segmentation. At its core, SCENE-Net employs Group Equivariant Non-Expansive Operators (GENEOs), a mechanism that leverages geometric priors for enhanced object identification. Our contribution extends the theoretical landscape of geometric learning, highlighting the utility of geometric observers as intrinsic biases in analyzing 3D environments. Through empirical testing and efficiency analysis, we demonstrate the performance of SCENE-Net in detecting power line supporting towers, a key application in forest fire prevention. Our results showcase the superior accuracy and resilience of our model to label noise, achieved with minimal computational resources—this instantiation of SCENE-Net has only eleven trainable parameters—thereby marking a significant step forward in trustworthy machine learning applied to 3D scene understanding. Our code is available in: https://github.com/dlavado/scene-net .en
dc.description.versionpublishersversion
dc.description.versionpublished
dc.format.extent16
dc.format.extent3677696
dc.identifier.doi10.1016/j.cviu.2025.104531
dc.identifier.issn1077-3142
dc.identifier.otherPURE: 151025210
dc.identifier.otherPURE UUID: 18e14123-33c9-48e8-bb35-a04fc9d3894f
dc.identifier.otherScopus: 105020956026
dc.identifier.otherWOS: 001607695500001
dc.identifier.otherORCID: /0000-0003-3071-6627/work/203688990
dc.identifier.urihttp://hdl.handle.net/10362/199726
dc.identifier.urlhttps://www.scopus.com/pages/publications/105020956026
dc.identifier.urlhttps://www.webofscience.com/wos/woscc/full-record/WOS:001607695500001
dc.language.isoeng
dc.peerreviewedyes
dc.subject3D semantic segmentation
dc.subjectGroup Equivariant Non-Expansive Operators
dc.subjectPoint clouds
dc.subjectPower grids
dc.subjectWhite-box models
dc.subjectSoftware
dc.subjectSignal Processing
dc.subjectComputer Vision and Pattern Recognition
dc.titleSCENE-Neten
dc.title.subtitleGeometric induction for interpretable and low-resource 3D pole detection with Group-Equivariant Non-Expansive Operatorsen
dc.typejournal article
degois.publication.firstPage1
degois.publication.lastPage16
degois.publication.titleComputer Vision and Image Understanding
degois.publication.volume262
dspace.entity.typePublication
rcaap.rightsopenAccess

Ficheiros

Principais
A mostrar 1 - 1 de 1
Miniatura
Nome:
Lavado_et_al._2026_..pdf
Tamanho:
3.51 MB
Formato:
Adobe Portable Document Format
Ver/Abrir

Coleções

Home collection (FCT)