Találati lista | Tudóstér

001-es BibID:	BIBFORM111611
035-os BibID:	(Scopus)85159117341 (WoS)000985898500001
Első szerző:	Abriha Dávid (geográfus)
Cím:	Strategies in training deep learning models to extract building from multisource images with small training sample sizes / Abriha, Dávid; Szabó, Szilárd
Dátum:	2023
ISSN:	1753-8947
Megjegyzések:	Building extraction from remote sensing data is an important topic inurban studies and the deep learning methods have an increasing roledue to their minimal requirements in training data to reach outstandingperformance. We aimed to investigate the original U-Net architecture'sefficiency in building segmentation with different number of trainingimages and the role of data augmentation based on multisourceremote sensing data with varying spatial and spectral resolutions(WorldView-2 [WV2], WorldView-3 [WV3] images and an aerialorthophoto [ORTHO]). When the trainings and predictions wereconducted on the same image, U-Net provided good results with veryfew training images (validation accuracies: 94-97%; 192 images).Combining the ORTHO's and WV2's training data for prediction on WV3provided poor results with low F1-score (0.184). However, the inclusionof only 48 WV3 training images significantly improved the F1-score(0.693), thus, most buildings were correctly identified. Accordingly,using only independent reference data (other than the target image) isnot enough to train an accurate model. In our case, the reference fromWW2 and ORTHO images did not provide an acceptable basis to train agood model, but a minimal number of training images from thetargeted WV3 improved the accuracy (F1-score: 69%).
Tárgyszavak:	Természettudományok Földtudományok idegen nyelvű folyóiratközlemény külföldi lapban folyóiratcikk building segmentation U-Net remote sensing urban analysis
Megjelenés:	International Journal of Digital Earth. - 16 : 1 (2023), p. 1707-1724. -
További szerzők:	Szabó Szilárd (1974-) (geográfus)
Pályázati támogatás:	TKP2020-NKA-04 Egyéb Kooperatív Doktori Program Egyéb NKFI K138079 Egyéb
Internet cím:	Szerző által megadott URL DOI Intézményi repozitóriumban (DEA) tárolt változat
Borító:
Saját polcon: