Segment Anything Model for Refugee-Dwelling Extraction with Few Samples from High-Resolution Satellite Imagery

Gao, Yunya; Zhao, Hui

doi:10.34726/5740

Record link:

http://hdl.handle.net/20.500.12708/194765
https://doi.org/10.34726/5740

Title:

Segment Anything Model for Refugee-Dwelling Extraction with Few Samples from High-Resolution Satellite Imagery

Citation:

Gao, Y., & Zhao, H. (2023). Segment Anything Model for Refugee-Dwelling Extraction with Few Samples from High-Resolution Satellite Imagery. In Proceedings of the 18th International Conference on Location Based Services (pp. 87–92). https://doi.org/10.34726/5740

reposiTUm DOI:

10.34726/5740

CatalogPlus:

AC17202176

Publication Type:

Inproceedings - Full-Paper Contribution

Language:

English

Authors:

Gao, Yunya
Zhao, Hui

Organisational Unit:

E000 - Technische Universität Wien

Published in:

Proceedings of the 18th International Conference on Location Based Services

DOI of the book:

10.34726/5400

Date (published):

2023

Event name:

18th International Conference on Location Based Services (LBS 2023)

Event date:

20-Nov-2023 - 22-Nov-2023

Event place:

Ghent, Belgium

Number of Pages:

Keywords:

Segment Anything Model; Dwelling Extraction

Abstract:

Customizing Segment Anything Model (SAM) has recently attracted considerable attention in remote sensing domains. This study explores the performance of SAM-Adapter in refugee-dwelling extraction in three different refugee camps from high-resolution satellite images. The findings indicate that with scarce sample data, SAM-Adapter marginally outperforms other semantic segmentation models. This underscores SAM's promising potential for building extraction tasks when data is limited.

Linked records in reposiTUm: is part of:

10.34726/5400

Research Areas:

Visual Computing and Human-Centered Technology: 100%

License:

CC BY 4.0