Remote Sensing Image Classification Via Vision Transformer and Transfer Learning

Abstract Aerial scene classification, which aims to automatically tag an aerial image with a specific semantic category, is a fundamental problem for understanding high-resolution remote sensing imagery. The classification of remote sensing image scenes can provide significant value, from forest fire monitoring to land use and land cover classification. From the first aerial photographs of the early 20th century to today's satellite imagery, the amount of remote sensing data has increased geometrically with higher resolution. The need to analyze this modern digital data has motivated research to accelerate the classification of remotely sensed images. Fortunately, the computer vision community has made great strides in classifying natural images. Transformers first applied to the field of natural language processing, is a type of deep neural network mainly based on the self-attention mechanism. Thanks to its strong representation capabilities, researchers are looking at ways to apply transformers to computer vision tasks. In a variety of visual benchmarks, transformer-based models perform similar to or better than other types of networks such as convolutional and recurrent networks. Given its high performance and less need for vision-specific inductive bias, the transformer is receiving more and more attention from the computer vision community. In this paper, we provide a systematic review of the Transfer Learning and Transformer techniques for scene classification using AID datasets. Both approaches give an accuracy of 80% and 84%, for the AID dataset. Keywords: remote sensing, vision transformers, transfer learning, classification accuracy