Abstract
Due to device limitations, small networks are necessary for some real-world scenarios, such as satellites and micro-robots. Therefore, the development of a network with both good performance and small size is an important area of research. Deep networks can learn well from large amounts of data, while manifold networks have outstanding feature representation at small sizes. In this paper, we propose an approach that exploits the advantages of deep networks and shallow Grassmannian manifold networks. Inspired by knowledge distillation, we use the information learned from convolutional neural networks to guide the training of the manifold networks. Our approach leads to a reduction in model size, which addresses the problem of deploying deep learning on resource-limited embedded devices. Finally, a series of experiments were conducted on four remote sensing scene classification datasets. The method in this paper improved the classification accuracy by 2.31% and 1.73% on the UC Merced Land Use and SIRIWHU datasets, respectively, and the experimental results demonstrate the effectiveness of our approach.
Highlights
In recent years, high-resolution remote sensing (HRRS) images have become more accessible with the development of satellite and remote sensing technologies, which provide detailed information on the land surface
In order to figure out the effect of different networks on knowledge distillation, we add the experiment of ResNet50 on this dataset
As can be seen from the table, the GrNet2B trained by knowledge distillation with the ResNet34 improves the accuracy by 1.09% compared with the GrNet2B trained by the ground truth alone
Summary
High-resolution remote sensing (HRRS) images have become more accessible with the development of satellite and remote sensing technologies, which provide detailed information on the land surface. Many remote sensing image tasks are developing rapidly, such as semantic segmentation [1,2], object detection [3]. Remote sensing image semantic segmentation has been widely used in various applications, such as natural resource protection, change detection [4]. Objects in high-resolution remote sensing images have rich details, such as geometry and structure, which bring more challenges to land use classification. The scene classification of optical remote sensing images can be divided into two categories, namely, methods based on artificially designed features and methods based on deep features. The hand-crafted features used for optical remote sensing image scene classification can be broadly classified into three categories, namely, spectral features, texture features, and structural features. The effectiveness of deep learning in remote sensing images has recently received a lot of attention. Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
