WA‐RSPrompter: Improved RSPrompter Combining Wavelet Transform and Self‐Attention Mechanism for Remote Sensing Image Instance Segmentation
IET Image Processing · 2026-01 · 20 (1)
DOI: 10.1049/ipr2.70443 ↗Kurz gefasst
Fachartikel in der Zeitschrift IET Image Processing, online erschienen am 04.08.2026. Verfasst von Jingjun Chen und 1 weiteren Autor:innen. Die Arbeit nennt 25 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Traditional deep learning approaches to remote sensing instance segmentation are inherently task‐specific and data‐intensive, which severely constrains their generalisability. Despite its strong generalisation on natural images, the Segment Anything Model (SAM) suffers from a notable domain gap in remote sensing applications. To bridge this gap, we propose WA‐RSPrompter, which extends RSPrompter with a discrete wavelet transform (DWT)‐based feature adapter and a spatial‐channel attention module. On the NWPU VHR‐10 dataset, WA‐RSPrompter achieves 0.575 mAP for instance segmentation, outperforming RSPrompter by 0.055 with only marginal increases in parameters and latency, and exceeding all compared conventional methods. Furthermore, without fine‐tuning on the HRSID dataset, WA‐RSPrompter surpasses RSPrompter by 0.077 mAP, demonstrating strong cross‐domain generalisation. Results show that WA‐RSPrompter delivers consistent accuracy gains over RSPrompter with minimal additional computational cost.
Zitieren
APA
Chen, J., & Chen, H. (2026). WA‐RSPrompter: Improved RSPrompter Combining Wavelet Transform and Self‐Attention Mechanism for Remote Sensing Image Instance Segmentation. IET Image Processing, 20(1). https://doi.org/10.1049/ipr2.70443
Vancouver
Chen J, Chen H. WA‐RSPrompter: Improved RSPrompter Combining Wavelet Transform and Self‐Attention Mechanism for Remote Sensing Image Instance Segmentation. IET Image Processing. 2026;20(1). doi:10.1049/ipr2.70443
BibTeX
@article{Chen2026,
author = {Chen, Jingjun and Chen, Hongshun},
title = {WA‐RSPrompter: Improved RSPrompter Combining Wavelet Transform and Self‐Attention Mechanism for Remote Sensing Image Instance Segmentation},
journal = {IET Image Processing},
year = {2026},
volume = {20},
number = {1},
doi = {10.1049/ipr2.70443}
}
EUVIMED