SAR-HUB: Pre-Training, Fine-Tuning, and Explaining
Since the current remote sensing pre-trained models trained on optical images are not as effective when applied to SAR image tasks, it is crucial to create sensor-specific SAR models with generalized feature representations and to demonstrate with evidence the limitations of optical pre-trained mode...
Main Authors: | , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2023-11-01
|
Series: | Remote Sensing |
Subjects: | |
Online Access: | https://www.mdpi.com/2072-4292/15/23/5534 |
_version_ | 1797399658605051904 |
---|---|
author | Haodong Yang Xinyue Kang Long Liu Yujiang Liu Zhongling Huang |
author_facet | Haodong Yang Xinyue Kang Long Liu Yujiang Liu Zhongling Huang |
author_sort | Haodong Yang |
collection | DOAJ |
description | Since the current remote sensing pre-trained models trained on optical images are not as effective when applied to SAR image tasks, it is crucial to create sensor-specific SAR models with generalized feature representations and to demonstrate with evidence the limitations of optical pre-trained models in downstream SAR tasks. The following aspects are the focus of this study: pre-training, fine-tuning, and explaining. First, we collect the current large-scale open-source SAR scene image classification datasets to pre-train a series of deep neural networks, including convolutional neural networks (CNNs) and vision transformers (ViT). A novel dynamic range adaptive enhancement method and a mini-batch class-balanced loss are proposed to tackle the challenges in SAR scene image classification. Second, the pre-trained models are transferred to various SAR downstream tasks compared with optical ones. Lastly, we propose a novel knowledge point interpretation method to reveal the benefits of the SAR pre-trained model with comprehensive and quantifiable explanations. This study is reproducible using open-source code and datasets, demonstrates generalization through extensive experiments on a variety of tasks, and is interpretable through qualitative and quantitative analyses. The codes and models are open source. |
first_indexed | 2024-03-09T01:44:18Z |
format | Article |
id | doaj.art-b146c31a62d14e7facd8cf1fd681edf9 |
institution | Directory Open Access Journal |
issn | 2072-4292 |
language | English |
last_indexed | 2024-03-09T01:44:18Z |
publishDate | 2023-11-01 |
publisher | MDPI AG |
record_format | Article |
series | Remote Sensing |
spelling | doaj.art-b146c31a62d14e7facd8cf1fd681edf92023-12-08T15:24:56ZengMDPI AGRemote Sensing2072-42922023-11-011523553410.3390/rs15235534SAR-HUB: Pre-Training, Fine-Tuning, and ExplainingHaodong Yang0Xinyue Kang1Long Liu2Yujiang Liu3Zhongling Huang4The BRain and Artificial INtelligence Lab (BRAIN LAB), School of Automation, Northwestern Polytechnical University, Xi’an 710072, ChinaSchool of Civil Aviation, Northwestern Polytechnical University, Xi’an 710072, ChinaThe BRain and Artificial INtelligence Lab (BRAIN LAB), School of Automation, Northwestern Polytechnical University, Xi’an 710072, ChinaThe BRain and Artificial INtelligence Lab (BRAIN LAB), School of Automation, Northwestern Polytechnical University, Xi’an 710072, ChinaThe BRain and Artificial INtelligence Lab (BRAIN LAB), School of Automation, Northwestern Polytechnical University, Xi’an 710072, ChinaSince the current remote sensing pre-trained models trained on optical images are not as effective when applied to SAR image tasks, it is crucial to create sensor-specific SAR models with generalized feature representations and to demonstrate with evidence the limitations of optical pre-trained models in downstream SAR tasks. The following aspects are the focus of this study: pre-training, fine-tuning, and explaining. First, we collect the current large-scale open-source SAR scene image classification datasets to pre-train a series of deep neural networks, including convolutional neural networks (CNNs) and vision transformers (ViT). A novel dynamic range adaptive enhancement method and a mini-batch class-balanced loss are proposed to tackle the challenges in SAR scene image classification. Second, the pre-trained models are transferred to various SAR downstream tasks compared with optical ones. Lastly, we propose a novel knowledge point interpretation method to reveal the benefits of the SAR pre-trained model with comprehensive and quantifiable explanations. This study is reproducible using open-source code and datasets, demonstrates generalization through extensive experiments on a variety of tasks, and is interpretable through qualitative and quantitative analyses. The codes and models are open source.https://www.mdpi.com/2072-4292/15/23/5534SAR image interpretationpre-trained modeltransfer learningexplainable artificial intelligence |
spellingShingle | Haodong Yang Xinyue Kang Long Liu Yujiang Liu Zhongling Huang SAR-HUB: Pre-Training, Fine-Tuning, and Explaining Remote Sensing SAR image interpretation pre-trained model transfer learning explainable artificial intelligence |
title | SAR-HUB: Pre-Training, Fine-Tuning, and Explaining |
title_full | SAR-HUB: Pre-Training, Fine-Tuning, and Explaining |
title_fullStr | SAR-HUB: Pre-Training, Fine-Tuning, and Explaining |
title_full_unstemmed | SAR-HUB: Pre-Training, Fine-Tuning, and Explaining |
title_short | SAR-HUB: Pre-Training, Fine-Tuning, and Explaining |
title_sort | sar hub pre training fine tuning and explaining |
topic | SAR image interpretation pre-trained model transfer learning explainable artificial intelligence |
url | https://www.mdpi.com/2072-4292/15/23/5534 |
work_keys_str_mv | AT haodongyang sarhubpretrainingfinetuningandexplaining AT xinyuekang sarhubpretrainingfinetuningandexplaining AT longliu sarhubpretrainingfinetuningandexplaining AT yujiangliu sarhubpretrainingfinetuningandexplaining AT zhonglinghuang sarhubpretrainingfinetuningandexplaining |