2022
DOI: 10.1109/tgrs.2021.3104032
|View full text |Cite
|
Sign up to set email alerts
|

Stagewise Unsupervised Domain Adaptation With Adversarial Self-Training for Road Segmentation of Remote-Sensing Images

Abstract: Semantic segmentation of remote sensing (RS) images is a challenging yet crucial task. While deep learning, particularly supervised learning with large-scale labeled datasets, has significantly advanced this field, acquiring high-quality labeled data is expensive and time-consuming. Additionally, variations in ground sampling distance, imaging equipment, and geographic differences cause domain shifts between datasets, which limit model performance across domains. Unsupervised domain adaptation (UDA) offers a s… Show more

Help me understand this report
View preprint versions

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1

Citation Types

0
34
0

Year Published

2022
2022
2024
2024

Publication Types

Select...
8
1

Relationship

0
9

Authors

Journals

citations
Cited by 111 publications
(34 citation statements)
references
References 45 publications
0
34
0
Order By: Relevance
“…2) Size of Spatial Neighborhood Patch: We evaluate the impact of the spatial neighborhood patch size on the classification accuracy in detail by setting the spatial neighborhood patch λ size range of [17][18][19][20][21][22][23][24][25][26][27][28][29][30][31] with a step size of 2. As shown in Fig.…”
Section: B Parameter Tuningmentioning
confidence: 99%
See 1 more Smart Citation
“…2) Size of Spatial Neighborhood Patch: We evaluate the impact of the spatial neighborhood patch size on the classification accuracy in detail by setting the spatial neighborhood patch λ size range of [17][18][19][20][21][22][23][24][25][26][27][28][29][30][31] with a step size of 2. As shown in Fig.…”
Section: B Parameter Tuningmentioning
confidence: 99%
“…Deep learning (DL) captures the advanced features of the original data adaptively through a hierarchical structure. As a powerful feature extraction tool [18], it has been successfully applied in the field of remote sensing [19]- [22]. Chen et al [23] proposed a combination of stacked autoencoder and principal component analysis (PCA) to extract spectral features.…”
mentioning
confidence: 99%
“…In addition, some new encoder-decoder networks [32], [33] are also proposed for road extraction. RoadDA [34] was present to address the DS issue in road extraction task via adversarial self-training.…”
Section: Related Workmentioning
confidence: 99%
“…Incorporating LiDAR reduces data-level per-pixel spectral shift caused by illumination variation, but there exist feature-level disparities between source and target scenes. With the aid of information from the source domain and the related features, the target domain can be adapted with the source domain and classified by domain adaptation (DA) techniques [27]- [29]. Inspired by the traditional DA method in machine learning, DA technology has been introduced to HSI classification to reduce the data shift and distribution bias [30]- [32].…”
Section: Introductionmentioning
confidence: 99%