MMCLS: An Efficient Vision Mamba-Based Framework for Martian Image Classification

Gazing upon the cosmos, humanity has never ceased its journey to explore the unknown realms. For Martian data analysis, efficient and stable image classification models are essential. Mamba, a novel variant of State Space Models (SSM), has recently emerged as a breakthrough in sequential data processing, demonstrating exceptional memory efficiency and computational speed. Building upon this advancement, we pioneer MMCLS, a novel method integrating vision Mamba module for Martian image classification. Comparative experiments demonstrate that MMCLS is both succinct and effective, achieving state-of-the-art performance on HiRISE and MSL datasets.