Over the past few decades, image classification has grown in popularity as a research topic. Applications for image classification include automated inspection, land use classification, object recognition, traffic monitoring, quality control, and the detection of cancer cells. To categorize land cover and land use, deep learning networks are used to analyze remote sensing images. Numerous deep neural networks are used for different image analysis applications, such as AlexNet, VGG16, GoogleNet and ResNet. In this work, we have used a modified VGG16 network to classify the EuroSat remote sensing image dataset. At first, we have used CNN based scratch networks like ResNet50, VGG16, EfficientNetB0 and DenseNet121 for classification and got better performance for VGG16 considering accuracy, overfitting and time of execution. Then we applied transfer learning for classification and still VGG16 shows good performance with reduced overfitting. At last, we have derived a modified VGG16 network for the classification of EuroSat dataset. In all the three cases VGG16 exhibits excellent performance with accuracy values of 99.05%, 89.31% and 99.19% respectively. By distinguishing between various land covers, the classification of remote sensing images helps us to make better use of the land area.