Keywords
Mạng EfficientNet
Dữ liệu vệ tinh Landcover
EfficientNet
Sattelite data
Working Languages
How to Cite
Abstract
The emergence of machine learning methods, including classification algorithms, segmentation techniques, and optimization algorithms, has significantly contributed to automating the image analysis process and improved accuracies in monitoring changes on the Earth's surface. In this study, the Euroset dataset was used to train various variants of EfficientNet models, applied to scene-based classification of land cover. The results show that the classification accuracy of EfficientNet B3, B4, and B5 reached respective values of 97.7%, 97.74%, 97.9%, and the models were used to classify several scenes in the Northeastern part of Vietnam. Some miss-match occurred and it may be attributed to the training dataset lacking certain land cover types that are prevalent in Vietnam