{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,20]],"date-time":"2026-03-20T12:13:44Z","timestamp":1774008824579,"version":"3.50.1"},"reference-count":48,"publisher":"MDPI AG","issue":"8","license":[{"start":{"date-parts":[[2022,4,10]],"date-time":"2022-04-10T00:00:00Z","timestamp":1649548800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"name":"STAEDLER Foundation","award":["NA"],"award-info":[{"award-number":["NA"]}]},{"name":"FAU Emerging Fields Initiative grant","award":["TAPE"],"award-info":[{"award-number":["TAPE"]}]},{"name":"DFG and FAU funding program &quot;Open Access Publication Funding&quot;","award":["NA"],"award-info":[{"award-number":["NA"]}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["Remote Sensing"],"abstract":"<jats:p>Screening clouds, shadows, and snow is a critical pre-processing step in many remote-sensing data processing pipelines that operate on satellite image data from polar and high mountain regions. We observe that the results of the state-of-the-art Fmask algorithm are not very accurate in polar and high mountain regions. Given the unavailability of large, labeled Sentinel-2 training datasets, we present a multi-stage self-training approach that trains a model to perform semantic segmentation on Sentinel-2 L1C images using the noisy Fmask labels for training and a small human-labeled dataset for validation. At each stage of the proposed iterative framework, we use a larger network architecture in comparison to the previous stage and train a new model. The trained model at each stage is then used to generate new training labels for a bigger dataset, which are used for training the model in the next stage. We select the best model during training in each stage by evaluating the multi-class segmentation metric, mean Intersection over Union (mIoU), on the small human-labeled validation dataset. This effectively helps to correct the noisy labels. Our model achieved an overall accuracy of 93% compared to the Fmask 4 and Sen2Cor 2.8, which achieved 75% and 76%, respectively. We believe our approach can also be adapted for other remote-sensing applications for training deep-learning models with imprecise labels.<\/jats:p>","DOI":"10.3390\/rs14081825","type":"journal-article","created":{"date-parts":[[2022,4,10]],"date-time":"2022-04-10T23:06:01Z","timestamp":1649631961000},"page":"1825","update-policy":"https:\/\/doi.org\/10.3390\/mdpi_crossmark_policy","source":"Crossref","is-referenced-by-count":17,"title":["A Self-Trained Model for Cloud, Shadow and Snow Detection in Sentinel-2 Images of Snow- and Ice-Covered Regions"],"prefix":"10.3390","volume":"14","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-5840-4935","authenticated-orcid":false,"given":"Kamal Gopikrishnan","family":"Nambiar","sequence":"first","affiliation":[{"name":"Chair of Multimedia Communications and Signal Processing, Friedrich-Alexander-Universit\u00e4t Erlangen-N\u00fcrnberg, 91054 Erlangen, Germany"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-3565-7788","authenticated-orcid":false,"given":"Veniamin I.","family":"Morgenshtern","sequence":"additional","affiliation":[{"name":"Chair of Multimedia Communications and Signal Processing, Friedrich-Alexander-Universit\u00e4t Erlangen-N\u00fcrnberg, 91054 Erlangen, Germany"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-7780-1525","authenticated-orcid":false,"given":"Philipp","family":"Hochreuther","sequence":"additional","affiliation":[{"name":"Institute of Geography, Friedrich-Alexander-Universit\u00e4t Erlangen-N\u00fcrnberg, 91054 Erlangen, Germany"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5055-8959","authenticated-orcid":false,"given":"Thorsten","family":"Seehaus","sequence":"additional","affiliation":[{"name":"Institute of Geography, Friedrich-Alexander-Universit\u00e4t Erlangen-N\u00fcrnberg, 91054 Erlangen, Germany"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5169-1567","authenticated-orcid":false,"given":"Matthias Holger","family":"Braun","sequence":"additional","affiliation":[{"name":"Institute of Geography, Friedrich-Alexander-Universit\u00e4t Erlangen-N\u00fcrnberg, 91054 Erlangen, Germany"}]}],"member":"1968","published-online":{"date-parts":[[2022,4,10]]},"reference":[{"key":"ref_1","doi-asserted-by":"crossref","first-page":"83","DOI":"10.1016\/j.rse.2011.10.028","article-title":"Object-based cloud and cloud shadow detection in Landsat imagery","volume":"118","author":"Zhu","year":"2012","journal-title":"Remote Sens. Environ."},{"key":"ref_2","doi-asserted-by":"crossref","first-page":"269","DOI":"10.1016\/j.rse.2014.12.014","article-title":"Improvement and expansion of the Fmask algorithm: Cloud, cloud shadow, and snow detection for Landsats 4\u20137, 8, and Sentinel 2 images","volume":"159","author":"Zhu","year":"2015","journal-title":"Remote Sens. Environ."},{"key":"ref_3","doi-asserted-by":"crossref","first-page":"111205","DOI":"10.1016\/j.rse.2019.05.024","article-title":"Fmask 4.0: Improved cloud and cloud shadow detection in Landsats 4\u20138 and Sentinel-2 imagery","volume":"231","author":"Qiu","year":"2019","journal-title":"Remote Sens. Environ."},{"key":"ref_4","unstructured":"Louis, J., Debaecker, V., Pflug, B., Main-Knorn, M., Bieniarz, J., Mueller-Wilm, U., Cadau, E., and Gascon, F. (2016, January 9\u201313). Sentinel-2 Sen2Cor: L2A processor for users. Proceedings of the ESA Living Planet Symposium Living Planet Symposium, Prague, Czech Republic."},{"key":"ref_5","doi-asserted-by":"crossref","first-page":"2662","DOI":"10.1109\/TGRS.2003.815404","article-title":"Multifeature texture analysis for the classification of clouds in satellite imagery","volume":"41","author":"Christodoulou","year":"2003","journal-title":"IEEE Trans. Geosci. Remote Sens."},{"key":"ref_6","doi-asserted-by":"crossref","first-page":"342","DOI":"10.1016\/j.rse.2017.01.026","article-title":"Multi-feature combined cloud and cloud shadow detection in GaoFen-1 wide field of view imagery","volume":"191","author":"Li","year":"2017","journal-title":"Remote Sens. Environ."},{"key":"ref_7","doi-asserted-by":"crossref","first-page":"7172","DOI":"10.1002\/2015JD024722","article-title":"A universal dynamic threshold cloud detection algorithm (UDTCDA) supported by a prior surface reflectance database","volume":"121","author":"Sun","year":"2016","journal-title":"J. Geophys. Res. Atmos."},{"key":"ref_8","doi-asserted-by":"crossref","first-page":"771","DOI":"10.5194\/isprs-archives-XLI-B8-771-2016","article-title":"An optional threshold with SVM cloud detection algorithm and DSP implementation","volume":"XLI-B8","author":"Zhou","year":"2016","journal-title":"Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci."},{"key":"ref_9","doi-asserted-by":"crossref","first-page":"5530","DOI":"10.1080\/01431161.2019.1580788","article-title":"Energy-based cloud detection in multispectral images based on the SVM technique","volume":"40","author":"Sui","year":"2019","journal-title":"Int. J. Remote Sens."},{"key":"ref_10","doi-asserted-by":"crossref","unstructured":"Hollstein, A., Segl, K., Guanter, L., Brell, M., and Enesco, M. (2016). Ready-to-use methods for the detection of clouds, cirrus, snow, shadow, water and clear sky pixels in Sentinel-2 MSI images. Remote Sens., 8.","DOI":"10.3390\/rs8080666"},{"key":"ref_11","doi-asserted-by":"crossref","first-page":"288","DOI":"10.1016\/j.asr.2018.04.030","article-title":"Introducing two random forest based methods for cloud detection in remote sensing images","volume":"62","author":"Ghasemian","year":"2018","journal-title":"Adv. Space Res."},{"key":"ref_12","doi-asserted-by":"crossref","first-page":"351","DOI":"10.1016\/j.isprsjprs.2008.12.007","article-title":"Use of Markov random fields for automatic cloud\/shadow detection on high resolution optical images","volume":"64","year":"2009","journal-title":"ISPRS J. Photogramm. Remote Sens."},{"key":"ref_13","doi-asserted-by":"crossref","first-page":"5100","DOI":"10.1109\/TGRS.2013.2286834","article-title":"A class of cloud detection algorithms based on a MAP-MRF approach in space and time","volume":"52","author":"Vivone","year":"2014","journal-title":"IEEE Trans. Geosci. Remote Sens."},{"key":"ref_14","first-page":"234","article-title":"U-Net: Convolutional networks for biomedical image segmentation","volume":"Volume 9351","author":"Ronneberger","year":"2015","journal-title":"Proceedings of the International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI)"},{"key":"ref_15","doi-asserted-by":"crossref","first-page":"70","DOI":"10.2352\/ISSN.2470-1173.2017.19.AVM-023","article-title":"Deep reinforcement learning framework for autonomous driving","volume":"2017","author":"Sallab","year":"2017","journal-title":"Electron. Imaging"},{"key":"ref_16","unstructured":"Sohn, K., Zhang, Z., Li, C.L., Zhang, H., Lee, C.Y., and Pfister, T. (2020). A simple semi-supervised learning framework for object detection. arXiv."},{"key":"ref_17","first-page":"3833","article-title":"Rethinking pre-training and self-training","volume":"Volume 33","author":"Larochelle","year":"2020","journal-title":"Advances in Neural Information Processing Systems"},{"key":"ref_18","doi-asserted-by":"crossref","unstructured":"Xie, Q., Luong, M.T., Hovy, E., and Le, Q.V. (2020, January 14\u201319). Self-training with noisy student improves ImageNet classification. Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA.","DOI":"10.1109\/CVPR42600.2020.01070"},{"key":"ref_19","doi-asserted-by":"crossref","first-page":"1","DOI":"10.5391\/IJFIS.2017.17.1.1","article-title":"Deep neural network self-training based on unsupervised learning and dropout","volume":"17","author":"Lee","year":"2017","journal-title":"Int. J. Fuzzy Log. Intell. Syst."},{"key":"ref_20","doi-asserted-by":"crossref","unstructured":"Babakhin, Y., Sanakoyeu, A., and Kitamura, H. (2019, January 10\u201313). Semi-supervised segmentation of salt bodies in seismic images using an ensemble of convolutional neural networks. Proceedings of the German Conference on Pattern Recognition (GCPR), Dortmund, Germany.","DOI":"10.1007\/978-3-030-33676-9_15"},{"key":"ref_21","doi-asserted-by":"crossref","unstructured":"Chen, L.C., Lopes, R.G., Cheng, B., Collins, M.D., Cubuk, E.D., Zoph, B., Adam, H., and Shlens, J. (2020, January 23\u201328). Naive-student: Leveraging semi-supervised learning in video sequences for urban scene segmentation. Proceedings of the European Conference on Computer Vision (ECCV), Glasgow, UK.","DOI":"10.1007\/978-3-030-58545-7_40"},{"key":"ref_22","unstructured":"Yilmaz, F.F., and Heckel, R. (2019). Image recognition from raw labels collected without annotators. arXiv."},{"key":"ref_23","doi-asserted-by":"crossref","first-page":"5449","DOI":"10.1080\/01431160903369642","article-title":"Automated masking of cloud and cloud shadow for forest change analysis using Landsat images","volume":"31","author":"Huang","year":"2010","journal-title":"Int. J. Remote Sens."},{"key":"ref_24","doi-asserted-by":"crossref","first-page":"1179","DOI":"10.14358\/PERS.72.10.1179","article-title":"Characterization of the Landsat-7 ETM+ automated cloud-cover assessment (ACCA) algorithm","volume":"72","author":"Irish","year":"2006","journal-title":"Photogramm. Eng. Remote Sens."},{"key":"ref_25","doi-asserted-by":"crossref","first-page":"379","DOI":"10.1016\/j.rse.2017.03.026","article-title":"Cloud detection algorithm comparison and validation for operational Landsat data products","volume":"194","author":"Foga","year":"2017","journal-title":"Remote Sens. Environ."},{"key":"ref_26","doi-asserted-by":"crossref","first-page":"307","DOI":"10.1016\/j.rse.2019.03.007","article-title":"Cloud and cloud shadow detection in Landsat imagery based on deep convolutional neural networks","volume":"225","author":"Chai","year":"2019","journal-title":"Remote Sens. Environ."},{"key":"ref_27","unstructured":"Xu, K., Guan, K., Peng, J., Luo, Y., and Wang, S. (2019). DeepMask: An algorithm for cloud and cloud shadow detection in optical satellite remote sensing images using deep residual network. arXiv."},{"key":"ref_28","doi-asserted-by":"crossref","first-page":"247","DOI":"10.1016\/j.rse.2019.03.039","article-title":"A cloud detection algorithm for satellite imagery based on deep learning","volume":"229","author":"Jeppesen","year":"2019","journal-title":"Remote Sens. Environ."},{"key":"ref_29","unstructured":"Mohajerani, S., and Saeedi, P. (August, January 28). Cloud-Net: An end-to-end cloud detection algorithm for Landsat 8 imagery. Proceedings of the IEEE International Geoscience and Remote Sensing Symposium (IGARSS), Yokohama, Japan."},{"key":"ref_30","doi-asserted-by":"crossref","first-page":"4062","DOI":"10.1109\/TGRS.2018.2889677","article-title":"Cloud detection in remote sensing images based on multiscale features-convolutional neural network","volume":"57","author":"Shao","year":"2019","journal-title":"IEEE Trans. Geosci. Remote Sens."},{"key":"ref_31","doi-asserted-by":"crossref","first-page":"1785","DOI":"10.1109\/LGRS.2017.2735801","article-title":"Distinguishing cloud and snow in satellite images via deep convolutional network","volume":"14","author":"Zhan","year":"2017","journal-title":"IEEE Geosci. Remote Sens. Lett."},{"key":"ref_32","doi-asserted-by":"crossref","first-page":"1600","DOI":"10.1109\/LGRS.2018.2846802","article-title":"Cloud and cloud shadow detection using multilevel feature fused segmentation network","volume":"15","author":"Yan","year":"2018","journal-title":"IEEE Geosci. Remote Sens. Lett."},{"key":"ref_33","first-page":"1","article-title":"Improving deep learning-based cloud detection for satellite images with attention mechanism","volume":"19","author":"Zhang","year":"2022","journal-title":"IEEE Geosci. Remote Sens. Lett."},{"key":"ref_34","doi-asserted-by":"crossref","unstructured":"Yu, J., Li, Y., Zheng, X., Zhong, Y., and He, P. (2020). An effective cloud detection method for Gaofen-5 images via deep learning. Remote Sens., 12.","DOI":"10.3390\/rs12132106"},{"key":"ref_35","first-page":"1","article-title":"DCNet: A deformable convolutional cloud detection network for remote sensing imagery","volume":"19","author":"Liu","year":"2022","journal-title":"IEEE Geosci. Remote Sens. Lett."},{"key":"ref_36","doi-asserted-by":"crossref","first-page":"112045","DOI":"10.1016\/j.rse.2020.112045","article-title":"Accurate cloud detection in high-resolution remote sensing imagery by weakly supervised deep learning","volume":"250","author":"Li","year":"2020","journal-title":"Remote Sens. Environ."},{"key":"ref_37","doi-asserted-by":"crossref","unstructured":"Liu, C.C., Zhang, Y.C., Chen, P.Y., Lai, C.C., Chen, Y.H., Cheng, J.H., and Ko, M.H. (2019). Clouds classification from Sentinel-2 imagery with deep residual learning and semantic image segmentation. Remote Sens., 11.","DOI":"10.3390\/rs11020119"},{"key":"ref_38","first-page":"1","article-title":"A lightweight deep learning-based cloud detection method for Sentinel-2A imagery fusing multiscale spectral and spatial features","volume":"60","author":"Li","year":"2022","journal-title":"IEEE Trans. Geosci. Remote Sens."},{"key":"ref_39","doi-asserted-by":"crossref","unstructured":"Hughes, M.J., and Kennedy, R. (2019). High-quality cloud masking of Landsat 8 imagery using convolutional neural networks. Remote Sens., 11.","DOI":"10.3390\/rs11212591"},{"key":"ref_40","unstructured":"ESA (2020, June 19). Sentinel-2 Spectral Band Information. Available online: https:\/\/sentinel.esa.int\/web\/sentinel\/user-guides\/sentinel-2-msi\/resolutions\/radiometric."},{"key":"ref_41","unstructured":"QGIS Development Team (2021). QGIS Geographic Information System, QGIS Association."},{"key":"ref_42","doi-asserted-by":"crossref","first-page":"107","DOI":"10.1016\/j.rse.2017.07.002","article-title":"Improving Fmask cloud and cloud shadow detection in mountainous area for Landsats 4\u20138 images","volume":"199","author":"Qiu","year":"2017","journal-title":"Remote Sens. Environ."},{"key":"ref_43","doi-asserted-by":"crossref","first-page":"127","DOI":"10.1016\/0034-4257(95)00137-P","article-title":"Development of methods for mapping global snow cover using moderate resolution imaging spectroradiometer data","volume":"54","author":"Hall","year":"1995","journal-title":"Remote Sens. Environ."},{"key":"ref_44","unstructured":"DeVries, T., and Taylor, G.W. (2017). Improved regularization of convolutional neural networks with cutout. arXiv."},{"key":"ref_45","doi-asserted-by":"crossref","unstructured":"Tompson, J., Goroshin, R., Jain, A., LeCun, Y., and Bregler, C. (2015, January 7\u201312). Efficient object localization using convolutional networks. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Boston, MA, USA.","DOI":"10.1109\/CVPR.2015.7298664"},{"key":"ref_46","doi-asserted-by":"crossref","unstructured":"Eigen, D., and Fergus, R. (2015, January 11\u201318). Predicting depth, surface normals and semantic labels with a common multi-scale convolutional architecture. Proceedings of the IEEE International Conference on Computer Vision (ICCV), Santiago, Chile.","DOI":"10.1109\/ICCV.2015.304"},{"key":"ref_47","unstructured":"Kampffmeyer, M., Salberg, A.B., and Jenssen, R. (July, January 29). Semantic segmentation of small objects and modeling of uncertainty in urban remote sensing images using deep convolutional neural networks. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, Las Vegas, NV, USA."},{"key":"ref_48","doi-asserted-by":"crossref","unstructured":"Martinuzzi, S., Gould, W.A., and Gonz\u00e1lez, O.M.R. (2007). Creating Cloud-Free Landsat ETM+ Data Sets in Tropical Landscapes: Cloud and Cloud-Shadow Removal, US Department of Agriculture, Forest Service, International Institute of Tropical Forestry. General Technical Report IITF-32.","DOI":"10.2737\/IITF-GTR-32"}],"container-title":["Remote Sensing"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.mdpi.com\/2072-4292\/14\/8\/1825\/pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,10,10]],"date-time":"2025-10-10T22:51:25Z","timestamp":1760136685000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.mdpi.com\/2072-4292\/14\/8\/1825"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2022,4,10]]},"references-count":48,"journal-issue":{"issue":"8","published-online":{"date-parts":[[2022,4]]}},"alternative-id":["rs14081825"],"URL":"https:\/\/doi.org\/10.3390\/rs14081825","relation":{},"ISSN":["2072-4292"],"issn-type":[{"value":"2072-4292","type":"electronic"}],"subject":[],"published":{"date-parts":[[2022,4,10]]}}}