A Robust and Low Complexity Deep Learning Model for Remote Sensing Image Classification

In this paper, we present a robust and low complexity deep learning model for Remote Sensing Image Classification (RSIC), the task of identifying the scene of a remote sensing image. In particular, we firstly evaluate different low complexity and benchmark deep neural networks: MobileNetV1, MobileNetV2, NASNetMobile, and EfficientNetB0, which present a number of trainable parameters lower than 5 Million (M) or occupy 20 MB memory. After indicating the best network architecture, we further improve the network performance by applying attention schemes to multiple feature maps extracted from middle layers of the network. To deal with the issue of increasing the model footprint due to using attention schemes, we apply the quantization technique to satisfy the maximum memory occupation of 20 MB. By conducting extensive experiments on the benchmark datasets NWPU-RESISC45, we achieve a robust and low-complexity model, which is very competitive to the state-of-the-art systems and potential for real-life applications on edge devices.