Image Classification of Satellite Using VGG16 Model
Ankita, Sonam Mittal
Abstract
The field of satellite image classification is constantly expanding and improving with the help of deep learning (DL) techniques. DL is a subset of machine learning that has made image classification more reliable than ever before. Satellite image classification, also known as remote sensing, is useful in many fields including disaster management, environmental issues, and disaster recovery. As a result, it has gained immense potential. This paper discusses the use of deep neural networks for image classification using the convolution neural network (CNN) VGG16. The dataset used in this study consisted of 5631 images with one main class called “data” and four sub-classes: cloudy, desert, green area, and water. VGG16 is a convolution neural network that uses multiple hidden layers to extract features and generate outputs. The VGG16 model showed high accuracy rates, with a train accuracy of 99.85% and a test accuracy of 99.91%.