Weproposeamethodforcontent-basedretrievingsolarmagnetograms.WeusetheSDOHelioseismicandMagneticImageroutputcollectedwithSunPyPyTorchlibraries.WecreateamathematicalrepresentationofthemagneticfieldregionsoftheSunintheformofavector.Thankstothissolutionwecancompareshortvectorsinsteadofcomparingfull-diskimages.Inordertodecreasetheretrievaltime,weusedafully-connectedau-toencoder,whichreducedthe256-elementdescriptortoa32-elementsemantichash.Theperformedexperimentsandcomparisonsprovedtheefficiencyoftheproposedapproach.Ourapproachhasthehighestprecisionvalueincomparisonwithotherstate-of-the-artmethods.Thepresentedmethodcanbeusednotonlyforsolarimageretrievalbutalsoforclassificationtasks.
Dettaglio pubblicazione
2022, JOURNAL OF ARTIFICIAL INTELLIGENCE AND SOFT COMPUTING RESEARCH, Pages 299-306
Semantic Hashing for Fast Solar Magnetogram Retrieval (01a Articolo in rivista)
Grycuk Rafał, Scherer Rafał, Marchlewska Alina, Napoli Christian
Gruppo di ricerca: Artificial Intelligence and Robotics
keywords