GLUSE: Enhanced Channel-Wise Adaptive Gated Linear Units SE for Onboard Satellite Earth Observation Image Classification

This study introduces ResNet-GLUSE, a lightweight ResNet variant enhanced with Gated Linear Unit-enhanced Squeeze-and-Excitation (GLUSE), an adaptive channel-wise attention mechanism. By integrating dynamic gating into the traditional SE framework, GLUSE improves feature recalibration while maintaining computational efficiency. Experiments on EuroSAT and PatternNet datasets confirm its effectiveness, achieving exceeding $\mathbf{\text{94\%}}\mathbf{\text{and}}\mathbf{\text{98\%}}\mathbf{\text{accuracy}}$, respectively. While $\mathbf{\text{MobileViT}}\mathbf{\text{achieves}}\mathbf{\text{99\%}}\mathbf{\text{accuracy}}$, ResNet-GLUSE offers $\mathbf{\text{33x}}\mathbf{\text{fewer}}\mathbf{\text{parameters,}}\mathbf{\text{27x}}\mathbf{\text{fewer}}\mathbf{\text{FLOPs,}}\mathbf{\text{33x}}\mathbf{\text{smaller}}\mathbf{\text{model}}\mathbf{\text{size}}\mathbf{\text{(MB),}}\approx \mathbf{\text{6x}}\mathbf{\text{lower}}\mathbf{\text{power}}\mathbf{\text{consumption}}\mathbf{\text{(W),}}\mathbf{\text{and}}\approx \mathbf{\text{3x}}\mathbf{\text{faster}}\mathbf{\text{inference}}\mathbf{\text{time}}\mathbf{\text{(s)}}$, making it significantly more efficient for onboard satellite deployment. Furthermore, due to its simplicity, ResNet-GLUSE can be easily mimicked for $\mathbf{\text{neuromorphic}}\mathbf{\text{computing}}$, enabling ultra-low power inference at just $\mathbf{\text{852.30}}\mathbf{\text{mW}}$ on Akida Brainchip. This balance between high accuracy and ultra-low resource consumption establishes ResNet-GLUSE as a practical solution for real-time Earth Observation (EO) tasks. Reproducible codes are available in our shared repository.