Recent advances in 3D Gaussian Splatting (3DGS) have significantly improved the speed and efficiency of novel view synthesis. Unlike neural implicit representations like Neural Radiance Fields (NeRF), which use neural networks to model 3D scenes, 3DGS uses Gaussian ellipsoids to represent scenes, enabling fast rendering and efficient downstream tasks such as geometry editing and physical simulation. 3DGS achieves real-time rendering at 30 FPS with comparable quality to NeRF, making it suitable for low-cost 3D content creation and real-time applications. 3DGS has been applied to various tasks, including 3D reconstruction, 3D editing, and physical simulation. For 3D reconstruction, methods like MipSplatting and MS3DGS address aliasing and improve rendering quality. Techniques such as VDGS and Scaffold-GS enhance view-dependent effects, while methods like Octree-GS and StopThePop improve dynamic scene reconstruction. Dynamic 3DGS extends to reconstructing scenes with time-varying attributes, using methods like 4D-GS and GauFRe for temporal consistency and deformation modeling. In terms of 3D editing, GaussianEditor and GaMeS enable geometry and appearance editing, while methods like TextureGS and 3DGM allow texture and lighting manipulation. Physical simulation methods like PhysGaussian and VR-GS enable realistic dynamic simulations, while methods like Spring-Gaus and Feature Splatting incorporate physical properties for accurate rendering. 3DGS has also been applied to applications such as segmentation, SLAM, and digital human creation. For segmentation, methods like LEGaussians and Gaussian Grouping integrate semantic features for accurate 3D scene understanding. In SLAM, methods like GS-SLAM and GaussianSplattingSLAM enable real-time localization and mapping. For digital humans, 3DGS has been used to create realistic avatars, with methods like GPS-Gaussian and GaussianAvatar achieving high-speed rendering and realistic animations. Overall, 3DGS provides an efficient and flexible representation for 3D scene modeling, enabling real-time rendering, dynamic reconstruction, and various downstream applications. Its explicit representation and efficient rendering make it a promising approach for future 3D content creation and real-time applications.Recent advances in 3D Gaussian Splatting (3DGS) have significantly improved the speed and efficiency of novel view synthesis. Unlike neural implicit representations like Neural Radiance Fields (NeRF), which use neural networks to model 3D scenes, 3DGS uses Gaussian ellipsoids to represent scenes, enabling fast rendering and efficient downstream tasks such as geometry editing and physical simulation. 3DGS achieves real-time rendering at 30 FPS with comparable quality to NeRF, making it suitable for low-cost 3D content creation and real-time applications. 3DGS has been applied to various tasks, including 3D reconstruction, 3D editing, and physical simulation. For 3D reconstruction, methods like MipSplatting and MS3DGS address aliasing and improve rendering quality. Techniques such as VDGS and Scaffold-GS enhance view-dependent effects, while methods like Octree-GS and StopThePop improve dynamic scene reconstruction. Dynamic 3DGS extends to reconstructing scenes with time-varying attributes, using methods like 4D-GS and GauFRe for temporal consistency and deformation modeling. In terms of 3D editing, GaussianEditor and GaMeS enable geometry and appearance editing, while methods like TextureGS and 3DGM allow texture and lighting manipulation. Physical simulation methods like PhysGaussian and VR-GS enable realistic dynamic simulations, while methods like Spring-Gaus and Feature Splatting incorporate physical properties for accurate rendering. 3DGS has also been applied to applications such as segmentation, SLAM, and digital human creation. For segmentation, methods like LEGaussians and Gaussian Grouping integrate semantic features for accurate 3D scene understanding. In SLAM, methods like GS-SLAM and GaussianSplattingSLAM enable real-time localization and mapping. For digital humans, 3DGS has been used to create realistic avatars, with methods like GPS-Gaussian and GaussianAvatar achieving high-speed rendering and realistic animations. Overall, 3DGS provides an efficient and flexible representation for 3D scene modeling, enabling real-time rendering, dynamic reconstruction, and various downstream applications. Its explicit representation and efficient rendering make it a promising approach for future 3D content creation and real-time applications.