A Network Architecture Search Remote Sensing Image Classification Method

The traditional deep convolutional neural network remote sensing image classification method requires trial and error many times, requiring a lot of time and computing resources. This paper proposes a remote sensing image classification method based on neural network architecture search. The method...

Full description

Saved in:
Bibliographic Details
Main Authors: JING Weipeng, ZHANG Mingwei, LIN Jingbo
Format: Article
Language:zho
Published: Harbin University of Science and Technology Publications 2021-02-01
Series:Journal of Harbin University of Science and Technology
Subjects:
Online Access:https://hlgxb.hrbust.edu.cn/#/digest?ArticleID=1914
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The traditional deep convolutional neural network remote sensing image classification method requires trial and error many times, requiring a lot of time and computing resources. This paper proposes a remote sensing image classification method based on neural network architecture search. The method first searches for an optimal cell, and then stacks the optimal cell in a predefined manner to obtain a target network. The method continues relaxation of the architecture represenation, so that the gradient descent method can be used to search in the discrete search space to achieve high efficiency. In order to improve its accuracy, we also retrain the target network on the training set The experimental results show that the classification accuracy of the model obtained by the test in the experimental test set is 88 57%, showing the high efficiency and high accuracy of the method.
ISSN:1007-2683