Title:	Attention-Aware DAE for Automated Solar Coronal Loop Segmentation
Authors:	Dhaubhadel, Prabal Man Lee, Jong Kwan Tian, Qing
Citation:	WSCG 2024: full papers proceedings: 32. International Conference in Central Europe on Computer Graphics, Visualization and Computer Vision, p. 67-76.
Issue Date:	2024
Publisher:	Václav Skala - UNION Agency
Document type:	konferenční příspěvek conferenceObject
URI:	http://hdl.handle.net/11025/57378
ISSN:	2464–4625 (online) 2464–4617 (print)
Keywords:	odšumovací autokodéry;blok pozornosti;segmentace solární koronální smyčky;vyplnění mezer smyčky
Keywords in different language:	denoising autoencoders;attention block;Solar Coronal Loop Segmentation;loop gap-filling
Abstract in different language:	This paper introduces an enhanced Denosing Autoencoder (DAE) model, incorporating a novel attention mecha nism, for the segmentation of solar coronal loops. This work is based on DAE framework to address the segmenta tion challenges posed by intricate structures of coronal loops which also appear with other solar features and image noises. Specifically, we introduce Encoding-Aware Decoding Attention (EADA) to all decoding stages of DAE, which resulted in improvement in coronal loop segmentation. Our models are validated through experiments on a synthetic image dataset of 11,000 images and a test dataset of 165 real coronal images of the NASA’s Solar Dy namics Observatory (SDO) satellite mission. Compared to the state-of-the-art coronal loop segmentation baseline, our attention-enhanced model results in better loop gap-filling and higher segmentation metrics (i.e., 3.6% increase in accuracy, 11.4% better recall and 5.6% higher precision).
Rights:	© Václav Skala - UNION Agency
Appears in Collections:	WSCG 2024: Full Papers Proceedings

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.