CDMamba: Remote Sensing Image Change Detection with Mamba CDMamba is a novel model that effectively combines global and local features for change detection (CD) tasks. The model addresses the challenge that current Mamba-based methods lack detailed features and struggle to achieve precise detection in dense prediction tasks. The Scaled Residual ConvMamba (SRCM) block is proposed to utilize Mamba's ability to extract global features and convolution to enhance local details, alleviating the issue of lacking detailed clues. Additionally, the Adaptive Global Local Guided Fusion (AGLGF) block is designed to facilitate bi-temporal interaction guided by global and local features, enabling the model to focus more on change regions and acquire discriminative differential features. Extensive experiments on three datasets (WHU-CD, LEVIR-CD, and LEVIR+CD) demonstrate that CDMamba outperforms current state-of-the-art methods. The model achieves state-of-the-art results in terms of F1 score and IoU, with improvements of 0.97%/1.21%, 0.98%/0.59%, and 2.10%/2.24% on the respective datasets. Qualitative results show that CDMamba provides more accurate detection results, particularly in complex scenes and small change areas. The model also demonstrates higher efficiency compared to Transformer-based methods, with fewer parameters and shorter training time. Ablation studies show that the SRCM and AGLGF modules significantly improve model performance. The SRCM module enhances the model's ability to capture global and local features, while the AGLGF module facilitates bi-temporal interaction and adaptive differential feature fusion. The model's performance is further improved by using non-saturating gate activation functions and expanding feature dimensions. The loss function coefficients are set to 0.5 for cross-entropy and dice loss to achieve balanced performance. In conclusion, CDMamba is a simple yet effective model that combines global and local information for change detection tasks. The model's performance is validated through extensive experiments on three public datasets, demonstrating its superiority over other state-of-the-art methods. Future work will explore the Mamba architecture for dense prediction tasks in remote sensing images by incorporating self-supervised learning methods.CDMamba: Remote Sensing Image Change Detection with Mamba CDMamba is a novel model that effectively combines global and local features for change detection (CD) tasks. The model addresses the challenge that current Mamba-based methods lack detailed features and struggle to achieve precise detection in dense prediction tasks. The Scaled Residual ConvMamba (SRCM) block is proposed to utilize Mamba's ability to extract global features and convolution to enhance local details, alleviating the issue of lacking detailed clues. Additionally, the Adaptive Global Local Guided Fusion (AGLGF) block is designed to facilitate bi-temporal interaction guided by global and local features, enabling the model to focus more on change regions and acquire discriminative differential features. Extensive experiments on three datasets (WHU-CD, LEVIR-CD, and LEVIR+CD) demonstrate that CDMamba outperforms current state-of-the-art methods. The model achieves state-of-the-art results in terms of F1 score and IoU, with improvements of 0.97%/1.21%, 0.98%/0.59%, and 2.10%/2.24% on the respective datasets. Qualitative results show that CDMamba provides more accurate detection results, particularly in complex scenes and small change areas. The model also demonstrates higher efficiency compared to Transformer-based methods, with fewer parameters and shorter training time. Ablation studies show that the SRCM and AGLGF modules significantly improve model performance. The SRCM module enhances the model's ability to capture global and local features, while the AGLGF module facilitates bi-temporal interaction and adaptive differential feature fusion. The model's performance is further improved by using non-saturating gate activation functions and expanding feature dimensions. The loss function coefficients are set to 0.5 for cross-entropy and dice loss to achieve balanced performance. In conclusion, CDMamba is a simple yet effective model that combines global and local information for change detection tasks. The model's performance is validated through extensive experiments on three public datasets, demonstrating its superiority over other state-of-the-art methods. Future work will explore the Mamba architecture for dense prediction tasks in remote sensing images by incorporating self-supervised learning methods.