Classification of satellite images with VGG19 and Convolutional Neural Network (CNN)

This work presents the application procedure VGG19, MobileNetV2, and CNN models for images Classification in remote sensing applications. This strategy employs pre-trained models, using transfer learning Feature extraction. The dataset includes classes like ’Cloudy,’ ’desert,’ ’green_field’ and ’water.’ Data augmentation techniques are involved to improve model generalization. The Models pass through training and evaluation with various considerations Setup, and class delivery are illustrated. Confusion matrix and precision-recall curves are analyzed. In addition, the Convolutional Neural Network (CNN) is used, focusing Data preparation, amplification and model design. The study presents the training, evaluation and comparison of two models version, shows effective image classification. The test result show a accuracy of $97.51 \%$ with consistently high Precision, recall, and F1-score across classes (’cloudy,’ ’desert,’ ’green_area,’ and ’water’), resulting in aggregates accuracy of $\mathbf{9 6 \%}$. These results underscore its effectiveness Transfer learning and CNN-based techniques in robust images classification is the critical step from data exploration to model deployment considerations are included, providing a comprehensive framework for remote sensing imagery classification.