{"title":"An optimized training approach for meteor detection with an attention mechanism to improve robustness on limited data","authors":"V.Y. Shirasuna, A.L.S. Gradvohl","doi":"10.1016/j.ascom.2023.100753","DOIUrl":null,"url":null,"abstract":"<div><p><span>Researchers have extensively used convolutional neural networks<span> to detect meteor falls on Earth. However, when dealing with limited available data, these networks may need more robustness to classify new real-world images correctly. This study proposes an optimized training approach of a pre-trained model with an attention mechanism<span> to achieve better generalization results in such a scenario. We compare two architectures, an optimized base model and another version with an attention mechanism. Furthermore, we present a new and publicly available optical meteor dataset that merges several public data sources. We used the merged dataset to train classification models combined with a stratified five-fold cross-validation strategy to determine the reliability of the prediction. The experimental results from both architectures showed good and similar performance. To further determine the best architecture, we performed an additional analysis with visual explanations in new observations. The architecture with an attention mechanism was the best model achieving a </span></span></span>false alarm ratio of 2.6% and an accuracy of 97%.</p></div>","PeriodicalId":48757,"journal":{"name":"Astronomy and Computing","volume":null,"pages":null},"PeriodicalIF":1.9000,"publicationDate":"2023-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Astronomy and Computing","FirstCategoryId":"101","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S2213133723000689","RegionNum":4,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"ASTRONOMY & ASTROPHYSICS","Score":null,"Total":0}
引用次数: 0
Abstract
Researchers have extensively used convolutional neural networks to detect meteor falls on Earth. However, when dealing with limited available data, these networks may need more robustness to classify new real-world images correctly. This study proposes an optimized training approach of a pre-trained model with an attention mechanism to achieve better generalization results in such a scenario. We compare two architectures, an optimized base model and another version with an attention mechanism. Furthermore, we present a new and publicly available optical meteor dataset that merges several public data sources. We used the merged dataset to train classification models combined with a stratified five-fold cross-validation strategy to determine the reliability of the prediction. The experimental results from both architectures showed good and similar performance. To further determine the best architecture, we performed an additional analysis with visual explanations in new observations. The architecture with an attention mechanism was the best model achieving a false alarm ratio of 2.6% and an accuracy of 97%.
Astronomy and ComputingASTRONOMY & ASTROPHYSICSCOMPUTER SCIENCE,-COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS
CiteScore
4.10
自引率
8.00%
发文量
67
期刊介绍:
Astronomy and Computing is a peer-reviewed journal that focuses on the broad area between astronomy, computer science and information technology. The journal aims to publish the work of scientists and (software) engineers in all aspects of astronomical computing, including the collection, analysis, reduction, visualisation, preservation and dissemination of data, and the development of astronomical software and simulations. The journal covers applications for academic computer science techniques to astronomy, as well as novel applications of information technologies within astronomy.