.png){kind=icon}
.png){kind=icon}
Return
MS 3 -Mamba: A multi-scale spectral–spatial hybrid Mamba for hyperspectral image classification
.png){kind=icon}
PRE
AI
H
B
G
H
F
DOI：10.1016/j.imavis.2026.106082.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• We propose MS³-Mamba, a novel hybrid framework for hyperspectral classification. • Adaptive spectral graph convolution captures long-range topological relations. • Grouped Transformer and bidirectional Mamba cascade models sequence evolution. • Hierarchical channel attention fusion effectively mitigates Hughes phenomenon. • MS³-Mamba achieves state-of-the-art accuracy with low computational overhead.
Journal
IF:
4.2
Papers:
4.0K
Citations:
6.7K
Organization
Cited Papers