A Deep Transfer Learning Framework Using Teacher–Student Structure for Land Cover Classification of Remote-Sensing Imagery
- 6 September 2023
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Geoscience and Remote Sensing Letters
- Vol. 21 (1545598X) , 1-5
Abstract
Deep learning techniques are widely used for land cover classification in remote sensing, primarily because they can effectively extract complex features from imagery data, which is essential for accurate land cover classification. However, the heterogeneity of data can pose a challenge to the generalizability of deep models. To address this issue, we propose a novel transfer learning framework for land cover classification using teacher-student structure. The proposed framework utilizes the knowledge acquired by the teacher model from large datasets to facilitate fine-tuning of the student model on small datasets, which prevents problems such as overfitting resulting from training large models on small datasets and inferior performance arising from the limited learning capacity of small models. To achieve this goal, we design a loss function base on central moment discrepancy and high-temperature softmax. In addition, we conducted experiments on three distinct pairs of datasets and found that our proposed framework outperforms both training the student model from scratch on the target domain and simply fine-tuning the teacher and student models from the source domain to the target domain. Specifically, we observed an average increase in mIoU of 9.9%, 2.1%, and 4.3%. These results demonstrate the effectiveness and generalizability of our proposed framework for remote sensing land cover classification.Keywords
Funding Information
- National Natural Science Foundation of China (42101346)
- Natural Resources Science and Technology Project of Hubei Province (ZRZY2023KJ10)
- Science and Technology Commission of Shanghai Municipality (22DZ1100800)
- China Postdoctoral Science Foundation (2020M680109)
This publication has 17 references indexed in Scilit:
- A Superpixel-Guided Unsupervised Fast Semantic Segmentation Method of Remote Sensing ImagesIEEE Geoscience and Remote Sensing Letters, 2022
- Land cover and fire accounts to support Indigenous land management: A pilot study of Yawuru CountryJournal of Environmental Management, 2022
- Global land use / land cover with Sentinel 2 and deep learning2021
- Understanding deep learning (still) requires rethinking generalizationCommunications of the ACM, 2021
- Spatio-temporal and cumulative effects of land use-land cover and climate change on two ecosystem services in the Colombian AndesScience of The Total Environment, 2019
- Robust unsupervised domain adaptation for neural networks via moment alignmentInformation Sciences, 2019
- Joint Deep Learning for land cover and land use classificationRemote Sensing of Environment, 2019
- Symmetrical Dense-Shortcut Deep Fully Convolutional Networks for Semantic Segmentation of Very-High-Resolution Remote Sensing ImagesIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2018
- Mind the Class Weight Bias: Weighted Maximum Mean Discrepancy for Unsupervised Domain Adaptation2017
- Pyramid Scene Parsing Network2017