Domain-Adversarial Training of Self-Attention-Based Networks for Land Cover Classification Using Multi-Temporal Sentinel-2 Satellite Imagery
The increasing availability of large-scale remote sensing labeled data has prompted researchers to develop increasingly precise and accurate data-driven models for land cover and crop classification (LC&CC). Moreover, with the introduction of self-attention and introspection mechanisms, deep lea...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2021-06-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/13/13/2564 |
| _version_ | 1827688343426564096 |
|---|---|
| author | Mauro Martini Vittorio Mazzia Aleem Khaliq Marcello Chiaberge |
| author_facet | Mauro Martini Vittorio Mazzia Aleem Khaliq Marcello Chiaberge |
| author_sort | Mauro Martini |
| collection | DOAJ |
| description | The increasing availability of large-scale remote sensing labeled data has prompted researchers to develop increasingly precise and accurate data-driven models for land cover and crop classification (LC&CC). Moreover, with the introduction of self-attention and introspection mechanisms, deep learning approaches have shown promising results in processing long temporal sequences in the multi-spectral domain with a contained computational request. Nevertheless, most practical applications cannot rely on labeled data, and in the field, surveys are a time-consuming solution that pose strict limitations to the number of collected samples. Moreover, atmospheric conditions and specific geographical region characteristics constitute a relevant domain gap that does not allow direct applicability of a trained model on the available dataset to the area of interest. In this paper, we investigate adversarial training of deep neural networks to bridge the domain discrepancy between distinct geographical zones. In particular, we perform a thorough analysis of domain adaptation applied to challenging multi-spectral, multi-temporal data, accurately highlighting the advantages of adapting state-of-the-art self-attention-based models for LC&CC to different target zones where labeled data are not available. Extensive experimentation demonstrated significant performance and generalization gain in applying domain-adversarial training to source and target regions with marked dissimilarities between the distribution of extracted features. |
| first_indexed | 2024-03-10T09:54:29Z |
| format | Article |
| id | doaj.art-1dfd99026908436bafaa4fd3cf13ad5d |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-03-10T09:54:29Z |
| publishDate | 2021-06-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-1dfd99026908436bafaa4fd3cf13ad5d2023-11-22T02:26:37ZengMDPI AGRemote Sensing2072-42922021-06-011313256410.3390/rs13132564Domain-Adversarial Training of Self-Attention-Based Networks for Land Cover Classification Using Multi-Temporal Sentinel-2 Satellite ImageryMauro Martini0Vittorio Mazzia1Aleem Khaliq2Marcello Chiaberge3Department of Electronics and Telecommunications, Politecnico di Torino, 10124 Turin, ItalyDepartment of Electronics and Telecommunications, Politecnico di Torino, 10124 Turin, ItalyPIC4SeR, Interdepartmental Centre for Service Robotics, Politecnico di Torino, 10129 Turin, ItalyDepartment of Electronics and Telecommunications, Politecnico di Torino, 10124 Turin, ItalyThe increasing availability of large-scale remote sensing labeled data has prompted researchers to develop increasingly precise and accurate data-driven models for land cover and crop classification (LC&CC). Moreover, with the introduction of self-attention and introspection mechanisms, deep learning approaches have shown promising results in processing long temporal sequences in the multi-spectral domain with a contained computational request. Nevertheless, most practical applications cannot rely on labeled data, and in the field, surveys are a time-consuming solution that pose strict limitations to the number of collected samples. Moreover, atmospheric conditions and specific geographical region characteristics constitute a relevant domain gap that does not allow direct applicability of a trained model on the available dataset to the area of interest. In this paper, we investigate adversarial training of deep neural networks to bridge the domain discrepancy between distinct geographical zones. In particular, we perform a thorough analysis of domain adaptation applied to challenging multi-spectral, multi-temporal data, accurately highlighting the advantages of adapting state-of-the-art self-attention-based models for LC&CC to different target zones where labeled data are not available. Extensive experimentation demonstrated significant performance and generalization gain in applying domain-adversarial training to source and target regions with marked dissimilarities between the distribution of extracted features.https://www.mdpi.com/2072-4292/13/13/2564domain adaptationTransformersdeep learningland cover classification |
| spellingShingle | Mauro Martini Vittorio Mazzia Aleem Khaliq Marcello Chiaberge Domain-Adversarial Training of Self-Attention-Based Networks for Land Cover Classification Using Multi-Temporal Sentinel-2 Satellite Imagery Remote Sensing domain adaptation Transformers deep learning land cover classification |
| title | Domain-Adversarial Training of Self-Attention-Based Networks for Land Cover Classification Using Multi-Temporal Sentinel-2 Satellite Imagery |
| title_full | Domain-Adversarial Training of Self-Attention-Based Networks for Land Cover Classification Using Multi-Temporal Sentinel-2 Satellite Imagery |
| title_fullStr | Domain-Adversarial Training of Self-Attention-Based Networks for Land Cover Classification Using Multi-Temporal Sentinel-2 Satellite Imagery |
| title_full_unstemmed | Domain-Adversarial Training of Self-Attention-Based Networks for Land Cover Classification Using Multi-Temporal Sentinel-2 Satellite Imagery |
| title_short | Domain-Adversarial Training of Self-Attention-Based Networks for Land Cover Classification Using Multi-Temporal Sentinel-2 Satellite Imagery |
| title_sort | domain adversarial training of self attention based networks for land cover classification using multi temporal sentinel 2 satellite imagery |
| topic | domain adaptation Transformers deep learning land cover classification |
| url | https://www.mdpi.com/2072-4292/13/13/2564 |
| work_keys_str_mv | AT mauromartini domainadversarialtrainingofselfattentionbasednetworksforlandcoverclassificationusingmultitemporalsentinel2satelliteimagery AT vittoriomazzia domainadversarialtrainingofselfattentionbasednetworksforlandcoverclassificationusingmultitemporalsentinel2satelliteimagery AT aleemkhaliq domainadversarialtrainingofselfattentionbasednetworksforlandcoverclassificationusingmultitemporalsentinel2satelliteimagery AT marcellochiaberge domainadversarialtrainingofselfattentionbasednetworksforlandcoverclassificationusingmultitemporalsentinel2satelliteimagery |