Research on Remote Sensing Image Classification Algorithm Based on EfficientNet

Abstract

Accurate classification of remote sensing images is important in remote sensing applications. In order to verify the efficiency and accuracy of efficientnet algorithm in remote sensing image classification, this paper classifies the UCMerced LandUse dataset based on EfficientNet. The experimental results show that compared with VGGNet, ResNet and MobileNet, the EfficientNet network introduces composite parameters and scales depth, width and resolution at the same time. The accuracy of EfficientV2-s in the verification set is 16.5%, 5.2%, 1.8% and 1.7% higher than that of VGG, MobileNetV2, ResNet34 and EfficientNet-b0, which shows the efficiency and accuracy of EfficientNet network in remote sensing image classification data set.