SEMI2I: Semantically Consistent Image-to-Image Translation for Domain Adaptation of Remote Sensing Data
- Resource Type
- Conference
- Authors
- Tasar, Onur; Happy, S L; Tarabalka, Yuliya; Alliez, Pierre
- Source
- IGARSS 2020 - 2020 IEEE International Geoscience and Remote Sensing Symposium Geoscience and Remote Sensing Symposium, IGARSS 2020 - 2020 IEEE International. :1837-1840 Sep, 2020
- Subject
- Aerospace
Computing and Processing
Geoscience
Photonics and Electrooptics
Signal Processing and Analysis
Training
Decoding
Training data
Image color analysis
Image segmentation
Image reconstruction
Roads
Domain adaptation
data augmentation
semantic segmentation
dense labeling
image-to-image translation
generative adversarial networks
GANs
- Language
- ISSN
- 2153-7003
Although convolutional neural networks have been proven to be an effective tool to generate high quality maps from remote sensing images, their performance significantly deteriorates when there exists a large domain shift between training and test data. To address this issue, we propose a new data augmentation approach that transfers the style of test data to training data using generative adversarial networks. Our semantic segmentation framework consists in first training a U-net from the real training data and then fine-tuning it on the test stylized fake training data generated by the proposed approach. Our experimental results prove that our framework outperforms the existing domain adaptation methods.