This paper proposes Remote Sensing Mamba (RSM), a novel model for dense prediction tasks in very-high-resolution (VHR) remote sensing images. RSM is designed to capture global context with linear complexity, enabling efficient processing of large VHR images without the need for patching. The model incorporates an omnidirectional selective scan module (OSSM) to globally model context in multiple directions, capturing large spatial features from various orientations. Extensive experiments on semantic segmentation and change detection tasks across various land covers demonstrate the effectiveness of RSM. The model achieves state-of-the-art performance on dense prediction tasks in VHR remote sensing images without fancy training strategies. RSM outperforms transformer-based models in terms of efficiency and accuracy on large remote sensing images. The model also performs better with larger image sizes on dense prediction tasks. The code is available at https://github.com/walking-shadow/Official_Remote_Sensing_Mamba. The paper also discusses the challenges of processing large VHR remote sensing images, the limitations of existing models, and the advantages of RSM in handling large images with rich contextual information. The model is evaluated on two tasks: semantic segmentation and change detection. The results show that RSM achieves the highest accuracy on both tasks. The paper also discusses the impact of image size and spatial resolution on dense prediction tasks in VHR remote sensing. The experiments show that larger images with higher spatial resolution lead to better performance. The model is capable of handling large VHR remote sensing images due to its linear complexity and global modeling capabilities. The paper concludes that RSM is a promising solution for dense prediction tasks in VHR remote sensing.This paper proposes Remote Sensing Mamba (RSM), a novel model for dense prediction tasks in very-high-resolution (VHR) remote sensing images. RSM is designed to capture global context with linear complexity, enabling efficient processing of large VHR images without the need for patching. The model incorporates an omnidirectional selective scan module (OSSM) to globally model context in multiple directions, capturing large spatial features from various orientations. Extensive experiments on semantic segmentation and change detection tasks across various land covers demonstrate the effectiveness of RSM. The model achieves state-of-the-art performance on dense prediction tasks in VHR remote sensing images without fancy training strategies. RSM outperforms transformer-based models in terms of efficiency and accuracy on large remote sensing images. The model also performs better with larger image sizes on dense prediction tasks. The code is available at https://github.com/walking-shadow/Official_Remote_Sensing_Mamba. The paper also discusses the challenges of processing large VHR remote sensing images, the limitations of existing models, and the advantages of RSM in handling large images with rich contextual information. The model is evaluated on two tasks: semantic segmentation and change detection. The results show that RSM achieves the highest accuracy on both tasks. The paper also discusses the impact of image size and spatial resolution on dense prediction tasks in VHR remote sensing. The experiments show that larger images with higher spatial resolution lead to better performance. The model is capable of handling large VHR remote sensing images due to its linear complexity and global modeling capabilities. The paper concludes that RSM is a promising solution for dense prediction tasks in VHR remote sensing.