The paper "RaDe-GS: Rasterizing Depth in Gaussian Splatting" introduces a novel method for computing depth and surface normal maps from 3D Gaussian splats, enhancing the accuracy of 3D shape reconstruction while maintaining the computational efficiency of Gaussian Splatting. The authors address the limitations of existing methods, which often suffer from poor shape accuracy due to the discrete and unstructured nature of Gaussian splats. Their approach involves rasterizing the depth evaluation in standard 3D Gaussian splats, allowing for precise surface reconstruction. The method is derived from general 3D Gaussian primitives and can be integrated into existing Gaussian Splatting-based methods. Experimental results on various datasets, including DTU, Tanks & Temples, and Mip-NeRF 360, demonstrate that RaDe-GS achieves superior performance in both novel view synthesis and 3D reconstruction compared to state-of-the-art implicit and explicit methods. The method also maintains high computational efficiency, making it a significant advancement in the field of Gaussian Splatting.The paper "RaDe-GS: Rasterizing Depth in Gaussian Splatting" introduces a novel method for computing depth and surface normal maps from 3D Gaussian splats, enhancing the accuracy of 3D shape reconstruction while maintaining the computational efficiency of Gaussian Splatting. The authors address the limitations of existing methods, which often suffer from poor shape accuracy due to the discrete and unstructured nature of Gaussian splats. Their approach involves rasterizing the depth evaluation in standard 3D Gaussian splats, allowing for precise surface reconstruction. The method is derived from general 3D Gaussian primitives and can be integrated into existing Gaussian Splatting-based methods. Experimental results on various datasets, including DTU, Tanks & Temples, and Mip-NeRF 360, demonstrate that RaDe-GS achieves superior performance in both novel view synthesis and 3D reconstruction compared to state-of-the-art implicit and explicit methods. The method also maintains high computational efficiency, making it a significant advancement in the field of Gaussian Splatting.