Smart City Gnosys
Smart city article details
| Title | Fattformer: Feature Attention Refinement Model For Semantic Segmentation In High-Resolution Remote Sensing Images |
|---|---|
| ID_Doc | 26188 |
| Authors | Tang S.; He L. |
| Year | 2025 |
| Published | Proceedings of SPIE - The International Society for Optical Engineering, 13506 |
| DOI | http://dx.doi.org/10.1117/12.3057505 |
| Abstract | High-resolution remote sensing images provide detailed surface information and are extensively utilized in urban planning, geographic monitoring, and smart city applications. Semantic segmentation, a crucial image analysis method, assigns pixel-level dense labels for scene understanding. However, this task faces challenges such as numerous small objects, complex shapes, detailed edge handling, and the effective fusion of multi-scale features. Existing methods often struggle to balance global information and local details, resulting in limited segmentation accuracy and efficiency. To address these challenges, we propose a novel network architecture named FAttFormer, which leverages the strengths of both Convolutional Neural Networks (CNNs) and Transformers. Within this architecture, we introduce two key modules: the Feature Attention Refinement Module (FARM) and the Multi-Scale Hybrid Attention (MSHA) module. The FAttFormer network is designed to significantly improve semantic segmentation accuracy and efficiency through multi-level and multi-scale feature fusion and refinement. Specifically, the FARM module focuses on effectively integrating encoder and decoder features to handle small objects and complex edges. The MSHA module enhances the ability to capture and integrate global and local information at different scales, further refining segmentation results. Extensive experiments on the ISPRS Vaihingen and Potsdam datasets demonstrate that FAttFormer outperforms existing models in segmenting impervious surfaces and buildings, showing significant improvements in accuracy and detail preservation. © 2025 SPIE. |
| Author Keywords | ISPRS Vaihingen dataset; remote sensing image; semantic segmentation; Transformer |
Similar Articles
| Id | Similarity | Authors | Title | Published |
|---|---|---|---|---|
8040  | 0.897 | Gonthina N.; Narasimha Prasad L.V. | An Enhanced Convolutional Neural Network Architecture For Semantic Segmentation In High-Resolution Remote Sensing Images | Discover Computing, 28, 1 (2025) |
| 17083 | 0.892 | Yan Y.; Li J.; Zhang J.; Wang L.; Zhuo L. | D2Sformer: Dual Attention-Dynamic Bidirectional Transformer For Semantic Segmentation Of Urban Remote Sensing Images | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 18 (2025) |
| 48265 | 0.87 | Arulananth T.S.; Kuppusamy P.G.; Ayyasamy R.K.; Alhashmi S.M.; Mahalakshmi M.; Vasanth K.; Chinnasamy P. | Semantic Segmentation Of Urban Environments: Leveraging U-Net Deep Learning Model For Cityscape Image Analysis | PLoS ONE, 19, 4 April (2024) |
| 35401 | 0.87 | Chen B.; Tong A.; Wang Y.; Zhang J.; Yang X.; Im S.-K. | Lkaffnet: A Novel Large-Kernel Attention Feature Fusion Network For Land Cover Segmentation | Sensors, 25, 1 (2025) |
| 48255 | 0.867 | Jia P.; Chen C.; Zhang D.; Sang Y.; Zhang L. | Semantic Segmentation Of Deep Learning Remote Sensing Images Based On Band Combination Principle: Application In Urban Planning And Land Use | Computer Communications, 217 (2024) |
| 48263 | 0.853 | Vetsa P.; Buddharaju A.; Dasari V.; Bogila Y.; Sundaresan A.A.; Paul M.B. | Semantic Segmentation Of Remote Sensing Images Of Urban Regions Using Deep Learning Methods | AIP Conference Proceedings, 3291, 1 (2025) |
| 13075 | 0.853 | Han Z.; Li X.; Wang X.; Wu Z.; Liu J. | Building Segmentation In Urban And Rural Areas With Mfa-Net: A Multidimensional Feature Adjustment Approach | Sensors, 25, 8 (2025) |