RS-DARTS: A Convolutional Neural Architecture Search for Remote Sensing Image Scene Classification
Due to the superiority of convolutional neural networks, many deep learning methods have been used in image classification. The enormous difference between natural images and remote sensing images makes it difficult to directly utilize or modify existing CNN models for remote sensing scene classific...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2021-12-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/14/1/141 |
| _version_ | 1797497704507506688 |
|---|---|
| author | Zhen Zhang Shanghao Liu Yang Zhang Wenbo Chen |
| author_facet | Zhen Zhang Shanghao Liu Yang Zhang Wenbo Chen |
| author_sort | Zhen Zhang |
| collection | DOAJ |
| description | Due to the superiority of convolutional neural networks, many deep learning methods have been used in image classification. The enormous difference between natural images and remote sensing images makes it difficult to directly utilize or modify existing CNN models for remote sensing scene classification tasks. In this article, a new paradigm is proposed that can automatically design a suitable CNN architecture for scene classification. A more efficient search framework, RS-DARTS, is adopted to find the optimal network architecture. This framework has two phases. In the search phase, some new strategies are presented, making the calculation process smoother, and better distinguishing the optimal and other operations. In addition, we added noise to suppress skip connections in order to close the gap between trained and validation processing and ensure classification accuracy. Moreover, a small part of the neural network is sampled to reduce the redundancy in exploring the network space and speed up the search processing. In the evaluation phase, the optimal cell architecture is stacked to construct the final network. Extensive experiments demonstrated the validity of the search strategy and the impressive classification performance of RS-DARTS on four public benchmark datasets. The proposed method showed more effectiveness than the manually designed CNN model and other methods of neural architecture search. Especially, in terms of search cost, RS-DARTS consumed less time than other NAS methods. |
| first_indexed | 2024-03-10T03:23:02Z |
| format | Article |
| id | doaj.art-5ad5175ba4e94d3080b2f74daed3f11e |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-03-10T03:23:02Z |
| publishDate | 2021-12-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-5ad5175ba4e94d3080b2f74daed3f11e2023-11-23T12:13:47ZengMDPI AGRemote Sensing2072-42922021-12-0114114110.3390/rs14010141RS-DARTS: A Convolutional Neural Architecture Search for Remote Sensing Image Scene ClassificationZhen Zhang0Shanghao Liu1Yang Zhang2Wenbo Chen3School of Information Science and Engineering, Lanzhou University, Lanzhou 730000, ChinaSchool of Physical Science and Technology, Lanzhou University, Lanzhou 730000, ChinaSchool of Information Science and Engineering, Lanzhou University, Lanzhou 730000, ChinaSchool of Information Science and Engineering, Lanzhou University, Lanzhou 730000, ChinaDue to the superiority of convolutional neural networks, many deep learning methods have been used in image classification. The enormous difference between natural images and remote sensing images makes it difficult to directly utilize or modify existing CNN models for remote sensing scene classification tasks. In this article, a new paradigm is proposed that can automatically design a suitable CNN architecture for scene classification. A more efficient search framework, RS-DARTS, is adopted to find the optimal network architecture. This framework has two phases. In the search phase, some new strategies are presented, making the calculation process smoother, and better distinguishing the optimal and other operations. In addition, we added noise to suppress skip connections in order to close the gap between trained and validation processing and ensure classification accuracy. Moreover, a small part of the neural network is sampled to reduce the redundancy in exploring the network space and speed up the search processing. In the evaluation phase, the optimal cell architecture is stacked to construct the final network. Extensive experiments demonstrated the validity of the search strategy and the impressive classification performance of RS-DARTS on four public benchmark datasets. The proposed method showed more effectiveness than the manually designed CNN model and other methods of neural architecture search. Especially, in terms of search cost, RS-DARTS consumed less time than other NAS methods.https://www.mdpi.com/2072-4292/14/1/141convolutional neural networks (CNNs)neural architecture search (NAS)remote sensing image scene classification |
| spellingShingle | Zhen Zhang Shanghao Liu Yang Zhang Wenbo Chen RS-DARTS: A Convolutional Neural Architecture Search for Remote Sensing Image Scene Classification Remote Sensing convolutional neural networks (CNNs) neural architecture search (NAS) remote sensing image scene classification |
| title | RS-DARTS: A Convolutional Neural Architecture Search for Remote Sensing Image Scene Classification |
| title_full | RS-DARTS: A Convolutional Neural Architecture Search for Remote Sensing Image Scene Classification |
| title_fullStr | RS-DARTS: A Convolutional Neural Architecture Search for Remote Sensing Image Scene Classification |
| title_full_unstemmed | RS-DARTS: A Convolutional Neural Architecture Search for Remote Sensing Image Scene Classification |
| title_short | RS-DARTS: A Convolutional Neural Architecture Search for Remote Sensing Image Scene Classification |
| title_sort | rs darts a convolutional neural architecture search for remote sensing image scene classification |
| topic | convolutional neural networks (CNNs) neural architecture search (NAS) remote sensing image scene classification |
| url | https://www.mdpi.com/2072-4292/14/1/141 |
| work_keys_str_mv | AT zhenzhang rsdartsaconvolutionalneuralarchitecturesearchforremotesensingimagesceneclassification AT shanghaoliu rsdartsaconvolutionalneuralarchitecturesearchforremotesensingimagesceneclassification AT yangzhang rsdartsaconvolutionalneuralarchitecturesearchforremotesensingimagesceneclassification AT wenbochen rsdartsaconvolutionalneuralarchitecturesearchforremotesensingimagesceneclassification |