The paper "RGBD GS-ICP SLAM" by Seongbo Ha, Jiung Yeon, and Hyeonwoo Yu from Sungkyunkwan University proposes a novel dense representation SLAM approach that integrates Generalized Iterative Closest Point (G-ICP) and 3D Gaussian Splatting (3DGS). This method leverages a single Gaussian map for both tracking and mapping, reducing redundant computations and achieving efficient system performance. The key contributions include: 1. **Real-time dense representation SLAM**: The proposed method achieves high speed (up to 107 FPS) and superior map quality. 2. **Active utilization of 3D information**: G-ICP is used for tracking, significantly reducing the time required for tracking processes. 3. **Reduction of computational cost**: Sharing covariances between G-ICP and 3DGS through scale alignment techniques minimizes unnecessary computations. 4. **Enhanced tracking accuracy and mapping quality**: Dynamic keyframe selection methods improve tracking accuracy and maintain consistent density of Gaussians in the map. The paper evaluates the proposed method on both synthetic and real-world datasets, demonstrating state-of-the-art performance in camera pose estimation, system speed, and map quality. The authors also conduct ablation studies to validate the effectiveness of their approach, showing that their method outperforms existing methods in various aspects. The code and video are available online for further research.The paper "RGBD GS-ICP SLAM" by Seongbo Ha, Jiung Yeon, and Hyeonwoo Yu from Sungkyunkwan University proposes a novel dense representation SLAM approach that integrates Generalized Iterative Closest Point (G-ICP) and 3D Gaussian Splatting (3DGS). This method leverages a single Gaussian map for both tracking and mapping, reducing redundant computations and achieving efficient system performance. The key contributions include: 1. **Real-time dense representation SLAM**: The proposed method achieves high speed (up to 107 FPS) and superior map quality. 2. **Active utilization of 3D information**: G-ICP is used for tracking, significantly reducing the time required for tracking processes. 3. **Reduction of computational cost**: Sharing covariances between G-ICP and 3DGS through scale alignment techniques minimizes unnecessary computations. 4. **Enhanced tracking accuracy and mapping quality**: Dynamic keyframe selection methods improve tracking accuracy and maintain consistent density of Gaussians in the map. The paper evaluates the proposed method on both synthetic and real-world datasets, demonstrating state-of-the-art performance in camera pose estimation, system speed, and map quality. The authors also conduct ablation studies to validate the effectiveness of their approach, showing that their method outperforms existing methods in various aspects. The code and video are available online for further research.