Remote sensing imagery is the key in fields such as environmental monitoring, urban planning, and disaster management. Deep learning methods have transformed image classification in remote sensing; hence, the accuracy and automation have been greatly enhanced. The main goal of this paper is to give an overview of the most important developments in the field, highlighting the methods such as auto-encoders, convolutional neural networks (CNN), generative adversarial networks (GAN), and the hybrid methods that come up with the combination of CNN and Transformers. Scene classification tasks are the focus of this work, particularly the use of efficient deep learning techniques and EfficientNetB0, which performed exceptionally. Among other models, EfficientNetB0 achieved an amazing accuracy of 99.50 on this task, overcoming the complexity of remote sensing imagery. Despite these successes, there are nonetheless challenges, including the tuning of hyperparameters, computational expense, and sensitivity to adversarial examples. For instance, the use of approaches, e.g., cross-domain transfer learning and task-specific application-specific models, will be necessary to address the problem of data scarcity and task-specific requirements. Potential future work should be to address computational issues, to do better than existing models, to improve robustness to object shape, texture and composition, and to be easier to interpret, all of which are critical steps towards realizing the complete potential of deep learning in classification.