Predicting Perceived Gloss: Do Weak Labels Suffice?

Guerrero-Viu, Julia (Universidad de Zaragoza) ; Subias, J. Daniel (Universidad de Zaragoza) ; Serrano, Ana (Universidad de Zaragoza) ; Storrs, Katherine R. ; Fleming, Roland W. ; Masia, Belen (Universidad de Zaragoza) ; Gutierrez, Diego (Universidad de Zaragoza)

Resumen: Estimating perceptual attributes of materials directly from images is a challenging task due to their complex, not fully‐understood interactions with external factors, such as geometry and lighting. Supervised deep learning models have recently been shown to outperform traditional approaches, but rely on large datasets of human‐annotated images for accurate perception predictions. Obtaining reliable annotations is a costly endeavor, aggravated by the limited ability of these models to generalise to different aspects of appearance. In this work, we show how a much smaller set of human annotations (“strong labels”) can be effectively augmented with automatically derived “weak labels” in the context of learning a low‐dimensional image‐computable gloss metric. We evaluate three alternative weak labels for predicting human gloss perception from limited annotated data. Incorporating weak labels enhances our gloss prediction beyond the current state of the art. Moreover, it enables a substantial reduction in human annotation costs without sacrificing accuracy, whether working with rendered images or real photographs.

Idioma: Inglés

DOI: 10.1111/cgf.15037

Año: 2024

Publicado en: Computer Graphics Forum 43, 2 (2024), e15037 [13 pp.]
ISSN: 0167-7055

Factor impacto JCR: 2.9 (2024)

Categ. JCR: COMPUTER SCIENCE, SOFTWARE ENGINEERING rank: 48 / 128 = 0.375 (2024) - Q2 - T2

Factor impacto SCIMAGO: 0.788 - Computer Graphics and Computer-Aided Design (Q1) - Computer Networks and Communications (Q2)

Financiación: info:eu-repo/grantAgreement/ES/DGA-CUS/702-2022
Financiación: info:eu-repo/grantAgreement/EC/HORIZON EUROPE/101098225/EU/Seeing Stuff: Perceiving Materials and their Properties/STUFF
Financiación: info:eu-repo/grantAgreement/EC/H2020/956585/EU/Predictive Rendering In Manufacture and Engineering/PRIME
Financiación: info:eu-repo/grantAgreement/ES/MCIU/FPU20-02340
Financiación: info:eu-repo/grantAgreement/ES/MICINN/PID2022-141766OB-I00

Tipo y forma: Article (Published version)

Área (Departamento): Área Lenguajes y Sistemas Inf. (Dpto. Informát.Ingenie.Sistms.)