KIT | KIT-Bibliothek | Impressum | Datenschutz

Gaussian Process Inspired Neural Networks for Spectral Unmixing Dataset Augmentation

Anastasiadis, Johannes ORCID iD icon; Heizmann, Michael


Hyperspectral imaging is increasingly used for product monitoring in industrial processes. Spectral unmixing is an important task in this context. As in many other areas of signal processing, neural networks also provide promising results for spectral unmixing. Unfortunately, it is very time-consuming to prepare labelled training data for the neural networks. To address this problem, this paper presents a method where small training datasets are augmented to improve spectral unmixing performance. Inspired by Gaussian processes, simple neural networks are trained which are capable of generating additional training data. These are similar to the original training data but cover areas in the continuous label space that are not covered by the original data.

Verlagsausgabe §
DOI: 10.5445/IR/1000141088
Veröffentlicht am 14.12.2021
Cover der Publikation
Zugehörige Institution(en) am KIT Institut für Industrielle Informationstechnik (IIIT)
Publikationstyp Proceedingsbeitrag
Publikationsdatum 27.10.2021
Sprache Englisch
Identifikator KITopen-ID: 1000141088
Erschienen in Artificial Intelligence. Ed.: K.-H. Schäfer
Veranstaltung The Upper Rhine Artificial Intelligence Symposium (UR-AI 2021), Kaiserslautern, Deutschland, 27.10.2021
Verlag Hochschule Kaiserslautern
Seiten 61-70
Externe Relationen Siehe auch
Schlagwörter spectral unmixing, spectral variability, data augmentation, neural, network, Gaussian process
Nachgewiesen in arXiv
KIT – Die Forschungsuniversität in der Helmholtz-Gemeinschaft
KITopen Landing Page