Weakly Supervised Domain Adversarial Neural Network for Deforestation Detection in Tropical Forests
Domain adaptation has proven to be suitable for alleviating domain discrepancies, which hinder the generalization capacity of classifiers. Among a few alternatives, domain adaptation techniques that align features in a domain-agnostic space through adversarial learning have been widely investigated....
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2023-01-01
|
| Series: | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10295968/ |
| _version_ | 1797448908176097280 |
|---|---|
| author | Pedro Juan Soto Vega Gilson Alexandre Ostwald Pedro da Costa Mabel Ximena Ortega Adarme Jose David Bermudez Castro Raul Queiroz Feitosa |
| author_facet | Pedro Juan Soto Vega Gilson Alexandre Ostwald Pedro da Costa Mabel Ximena Ortega Adarme Jose David Bermudez Castro Raul Queiroz Feitosa |
| author_sort | Pedro Juan Soto Vega |
| collection | DOAJ |
| description | Domain adaptation has proven to be suitable for alleviating domain discrepancies, which hinder the generalization capacity of classifiers. Among a few alternatives, domain adaptation techniques that align features in a domain-agnostic space through adversarial learning have been widely investigated. Nevertheless, such an approach often implies the deterioration of feature discriminability as a side effect of the adversarial alignment, which does not take into consideration class labels of the target domain samples. We advocate that weakly-supervised learning can mitigate that problem, as noisy labels for the target domain samples may serve to sustain class discriminability during the feature alignment procedure. Therefore, in this work we propose a weakly-supervised, adversarial domain adaptation method for a change detection task based on the Domain Adversarial Neural Network (DANN) strategy. We assessed the performance of the proposed method on a deforestation detection application, conducting experiments on sites of the Amazon and Cerrado biomes using Landsat-8 images. The results showed that the inclusion of weak supervision in the domain adaptation procedure provided higher accuracies than the original DANN strategy, which did not prescribe any supervision for the selection of target domain samples in training. On average, the Average Precision and F1-score values increased by 10.1\% and 12.6\% respectively with the use of the proposed method. Additionally, our method achieved compatible performances with the ones obtained by state-of-the-art domain adaptation methods. To the best of our knowledge, the proposed method is the first weakly-supervised domain adaptation strategy conceived for deforestation detection and, in general, for change detection. |
| first_indexed | 2024-03-09T14:17:15Z |
| format | Article |
| id | doaj.art-e5d40fc1c31047f5bd2810759ee1d29a |
| institution | Directory Open Access Journal |
| issn | 2151-1535 |
| language | English |
| last_indexed | 2024-03-09T14:17:15Z |
| publishDate | 2023-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
| spelling | doaj.art-e5d40fc1c31047f5bd2810759ee1d29a2023-11-29T00:00:22ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing2151-15352023-01-0116102641027810.1109/JSTARS.2023.332757310295968Weakly Supervised Domain Adversarial Neural Network for Deforestation Detection in Tropical ForestsPedro Juan Soto Vega0https://orcid.org/0000-0001-5396-8531Gilson Alexandre Ostwald Pedro da Costa1https://orcid.org/0000-0001-7341-9118Mabel Ximena Ortega Adarme2https://orcid.org/0000-0002-4106-0291Jose David Bermudez Castro3https://orcid.org/0000-0002-4516-5787Raul Queiroz Feitosa4https://orcid.org/0000-0001-8344-5096LaTIM, INSERM, UMR 1101, University Brest, Brest, FranceState University of Rio de Janeiro, Rio de Janeiro, BrazilPontifical Catholic University of Rio de Janairo, Rio de Janeiro, BrazilMcMaster University, Hamilton, ON, CanadaPontifical Catholic University of Rio de Janairo, Rio de Janeiro, BrazilDomain adaptation has proven to be suitable for alleviating domain discrepancies, which hinder the generalization capacity of classifiers. Among a few alternatives, domain adaptation techniques that align features in a domain-agnostic space through adversarial learning have been widely investigated. Nevertheless, such an approach often implies the deterioration of feature discriminability as a side effect of the adversarial alignment, which does not take into consideration class labels of the target domain samples. We advocate that weakly-supervised learning can mitigate that problem, as noisy labels for the target domain samples may serve to sustain class discriminability during the feature alignment procedure. Therefore, in this work we propose a weakly-supervised, adversarial domain adaptation method for a change detection task based on the Domain Adversarial Neural Network (DANN) strategy. We assessed the performance of the proposed method on a deforestation detection application, conducting experiments on sites of the Amazon and Cerrado biomes using Landsat-8 images. The results showed that the inclusion of weak supervision in the domain adaptation procedure provided higher accuracies than the original DANN strategy, which did not prescribe any supervision for the selection of target domain samples in training. On average, the Average Precision and F1-score values increased by 10.1\% and 12.6\% respectively with the use of the proposed method. Additionally, our method achieved compatible performances with the ones obtained by state-of-the-art domain adaptation methods. To the best of our knowledge, the proposed method is the first weakly-supervised domain adaptation strategy conceived for deforestation detection and, in general, for change detection.https://ieeexplore.ieee.org/document/10295968/Change detection (CD)deep learning (DL)deforestation detectiondomain adaptation (DA)weak supervision |
| spellingShingle | Pedro Juan Soto Vega Gilson Alexandre Ostwald Pedro da Costa Mabel Ximena Ortega Adarme Jose David Bermudez Castro Raul Queiroz Feitosa Weakly Supervised Domain Adversarial Neural Network for Deforestation Detection in Tropical Forests IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing Change detection (CD) deep learning (DL) deforestation detection domain adaptation (DA) weak supervision |
| title | Weakly Supervised Domain Adversarial Neural Network for Deforestation Detection in Tropical Forests |
| title_full | Weakly Supervised Domain Adversarial Neural Network for Deforestation Detection in Tropical Forests |
| title_fullStr | Weakly Supervised Domain Adversarial Neural Network for Deforestation Detection in Tropical Forests |
| title_full_unstemmed | Weakly Supervised Domain Adversarial Neural Network for Deforestation Detection in Tropical Forests |
| title_short | Weakly Supervised Domain Adversarial Neural Network for Deforestation Detection in Tropical Forests |
| title_sort | weakly supervised domain adversarial neural network for deforestation detection in tropical forests |
| topic | Change detection (CD) deep learning (DL) deforestation detection domain adaptation (DA) weak supervision |
| url | https://ieeexplore.ieee.org/document/10295968/ |
| work_keys_str_mv | AT pedrojuansotovega weaklysuperviseddomainadversarialneuralnetworkfordeforestationdetectionintropicalforests AT gilsonalexandreostwaldpedrodacosta weaklysuperviseddomainadversarialneuralnetworkfordeforestationdetectionintropicalforests AT mabelximenaortegaadarme weaklysuperviseddomainadversarialneuralnetworkfordeforestationdetectionintropicalforests AT josedavidbermudezcastro weaklysuperviseddomainadversarialneuralnetworkfordeforestationdetectionintropicalforests AT raulqueirozfeitosa weaklysuperviseddomainadversarialneuralnetworkfordeforestationdetectionintropicalforests |