Contrastive Dual-Pool Feature Adaption for Domain Incremental Remote Sensing Scene Classification
Remote sensing image classification has achieved remarkable success in environmental monitoring and urban planning using deep neural networks (DNNs). However, the performance of these models is significantly impacted by domain shifts due to seasonal changes, varying atmospheric conditions, and diffe...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-01-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/17/2/308 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832587609104187392 |
|---|---|
| author | Yingzhao Shao Yunsong Li Xiaodong Han |
| author_facet | Yingzhao Shao Yunsong Li Xiaodong Han |
| author_sort | Yingzhao Shao |
| collection | DOAJ |
| description | Remote sensing image classification has achieved remarkable success in environmental monitoring and urban planning using deep neural networks (DNNs). However, the performance of these models is significantly impacted by domain shifts due to seasonal changes, varying atmospheric conditions, and different geographical locations. Existing solutions, including rehearsal-based and prompt-based methods, face limitations such as data privacy concerns, high computational overhead, and unreliable feature embeddings due to domain gaps. To address these challenges, we propose DACL (dual-pool architecture with contrastive learning), a novel framework for domain incremental learning in remote sensing image classification. DACL introduces three key components: (1) a dual-pool architecture comprising a prompt pool for domain-specific tokens and an adapter pool for feature adaptation, enabling efficient domain-specific feature extraction; (2) a dual loss mechanism that combines image-attracting loss and text-separating loss to enhance intra-domain feature discrimination while maintaining clear class boundaries; and (3) a K-means-based domain selector that efficiently matches unknown domain features with existing domain representations using cosine similarity. Our approach eliminates the need for storing historical data while maintaining minimal computational overhead. Extensive experiments on six widely used datasets demonstrate that DACL consistently outperforms state-of-the-art methods in domain incremental learning for remote sensing image classification scenarios, achieving an average accuracy improvement of 4.07% over the best baseline method. |
| format | Article |
| id | doaj-art-f19888296c29465e944425766131736a |
| institution | Kabale University |
| issn | 2072-4292 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj-art-f19888296c29465e944425766131736a2025-01-24T13:48:04ZengMDPI AGRemote Sensing2072-42922025-01-0117230810.3390/rs17020308Contrastive Dual-Pool Feature Adaption for Domain Incremental Remote Sensing Scene ClassificationYingzhao Shao0Yunsong Li1Xiaodong Han2State Key Laboratory of Integrated Services Networks, Xidian University, Xi’an 710000, ChinaState Key Laboratory of Integrated Services Networks, Xidian University, Xi’an 710000, ChinaChina Academy of Space Technology, Xi’an 710100, ChinaRemote sensing image classification has achieved remarkable success in environmental monitoring and urban planning using deep neural networks (DNNs). However, the performance of these models is significantly impacted by domain shifts due to seasonal changes, varying atmospheric conditions, and different geographical locations. Existing solutions, including rehearsal-based and prompt-based methods, face limitations such as data privacy concerns, high computational overhead, and unreliable feature embeddings due to domain gaps. To address these challenges, we propose DACL (dual-pool architecture with contrastive learning), a novel framework for domain incremental learning in remote sensing image classification. DACL introduces three key components: (1) a dual-pool architecture comprising a prompt pool for domain-specific tokens and an adapter pool for feature adaptation, enabling efficient domain-specific feature extraction; (2) a dual loss mechanism that combines image-attracting loss and text-separating loss to enhance intra-domain feature discrimination while maintaining clear class boundaries; and (3) a K-means-based domain selector that efficiently matches unknown domain features with existing domain representations using cosine similarity. Our approach eliminates the need for storing historical data while maintaining minimal computational overhead. Extensive experiments on six widely used datasets demonstrate that DACL consistently outperforms state-of-the-art methods in domain incremental learning for remote sensing image classification scenarios, achieving an average accuracy improvement of 4.07% over the best baseline method.https://www.mdpi.com/2072-4292/17/2/308domain incremental learningvision language modelimage cluster and text separating |
| spellingShingle | Yingzhao Shao Yunsong Li Xiaodong Han Contrastive Dual-Pool Feature Adaption for Domain Incremental Remote Sensing Scene Classification Remote Sensing domain incremental learning vision language model image cluster and text separating |
| title | Contrastive Dual-Pool Feature Adaption for Domain Incremental Remote Sensing Scene Classification |
| title_full | Contrastive Dual-Pool Feature Adaption for Domain Incremental Remote Sensing Scene Classification |
| title_fullStr | Contrastive Dual-Pool Feature Adaption for Domain Incremental Remote Sensing Scene Classification |
| title_full_unstemmed | Contrastive Dual-Pool Feature Adaption for Domain Incremental Remote Sensing Scene Classification |
| title_short | Contrastive Dual-Pool Feature Adaption for Domain Incremental Remote Sensing Scene Classification |
| title_sort | contrastive dual pool feature adaption for domain incremental remote sensing scene classification |
| topic | domain incremental learning vision language model image cluster and text separating |
| url | https://www.mdpi.com/2072-4292/17/2/308 |
| work_keys_str_mv | AT yingzhaoshao contrastivedualpoolfeatureadaptionfordomainincrementalremotesensingsceneclassification AT yunsongli contrastivedualpoolfeatureadaptionfordomainincrementalremotesensingsceneclassification AT xiaodonghan contrastivedualpoolfeatureadaptionfordomainincrementalremotesensingsceneclassification |