A MobileNetV2 model of transfer learning is employed for remote sensing image classification

Abstract

Remote sensing image classification is an important and complicated problem in deep learning field. In order to achieve better classification effect, predecessors have proposed different kinds of models. A transfer learning-based MobileNetV2 model is suggested in this paper. First, we take UC-Merced dataset as input and introduce lightweight network MobileNetV2 for scene identification. Secondly, we combined the transfer learning training method and pre-trained the MobileNetV2 model to realize the high-performance classification of deeper remote sensing images. The UC- Merced dataset was classified with a model that achieved 91.43% accuracy,91.00% kappa index, and 91.50% F1-score. These results demonstrate the model's impressive performance in remote sensing image classification and its potential for practical scene identification applications.