Conference Papers Year : 2022

Deep Complex Convolutional Neural Networks for Remote Sensing Image Classification

Abstract

At present, the neural network is often based on the real field of operation, research shows that, compared with the real field, the complex has incomparable advantages in the field of image processing, such as the complex represents more information, such as the phase information and modulus value, which play a great role in some fields. To take full advantage of complex data, This paper mainly studies CNN network, and through complex value processing, and get Complex Convolutional Neural Networks(CCN), complete the construction of complex convolution neural network. In order to study complex neural network, we start from two aspects, one is convolution operation, the other is network construction. In this paper, we use ENet as the basic structure of the model, replace the convolutional structure, pooling structure, and BatchNorm structure with the complex form, use it in the Flevoland dataset, and get a good test results.
Fichier principal
Vignette du fichier
537972_1_En_27_Chapter.pdf (1.84 Mo) Télécharger le fichier
Origin Files produced by the author(s)

Dates and versions

hal-04666442 , version 1 (01-08-2024)

Licence

Identifiers

Cite

Lingling Li, Yukai Sun, Fuhai Ma, Jingjing Ma, Licheng Jiao, et al.. Deep Complex Convolutional Neural Networks for Remote Sensing Image Classification. 5th International Conference on Intelligence Science (ICIS), Oct 2022, Xi'an, China. pp.252-259, ⟨10.1007/978-3-031-14903-0_27⟩. ⟨hal-04666442⟩
20 View
1 Download

Altmetric

Share

More