This paper provides a comprehensive review of Gaussian Splatting, a state-of-the-art technique for 3D reconstruction and novel view synthesis (NVS). The authors highlight the challenges and advancements in 3D reconstruction, emphasizing the role of learning-based methods in improving performance and introducing novel capabilities. Gaussian Splatting, introduced in June 2023, is a novel approach that generates 3D objects from 2D images by iteratively refining multiple Gaussians. The paper covers input types, model structures, output representations, and training strategies, while also discussing unresolved challenges and future directions. The introduction section explains the importance of 3D reconstruction and NVS in various fields, such as computer graphics, robotics, and virtual reality. It discusses traditional and novel approaches to 3D data representation, including point clouds, meshes, and voxels, and highlights the advantages of Gaussian Splatting over traditional methods. The paper then delves into the mathematical representation and rendering process of Gaussian Splatting, detailing the projection of Gaussians, depth compositing, and quality assessment metrics like PSNR, SSIM, and LPIPS. It also explores the functional advancements in Gaussian Splatting, including dynamic and deformation capabilities, text-based generation, and optimization for faster training and inference. The review concludes with a discussion on specific applications and future research directions, emphasizing the versatility and potential of Gaussian Splatting in addressing various challenges in 3D reconstruction and NVS. The authors aim to provide a thorough overview of the latest advancements in Gaussian Splatting, highlighting its significance in the rapidly evolving field of 3D computer vision.This paper provides a comprehensive review of Gaussian Splatting, a state-of-the-art technique for 3D reconstruction and novel view synthesis (NVS). The authors highlight the challenges and advancements in 3D reconstruction, emphasizing the role of learning-based methods in improving performance and introducing novel capabilities. Gaussian Splatting, introduced in June 2023, is a novel approach that generates 3D objects from 2D images by iteratively refining multiple Gaussians. The paper covers input types, model structures, output representations, and training strategies, while also discussing unresolved challenges and future directions. The introduction section explains the importance of 3D reconstruction and NVS in various fields, such as computer graphics, robotics, and virtual reality. It discusses traditional and novel approaches to 3D data representation, including point clouds, meshes, and voxels, and highlights the advantages of Gaussian Splatting over traditional methods. The paper then delves into the mathematical representation and rendering process of Gaussian Splatting, detailing the projection of Gaussians, depth compositing, and quality assessment metrics like PSNR, SSIM, and LPIPS. It also explores the functional advancements in Gaussian Splatting, including dynamic and deformation capabilities, text-based generation, and optimization for faster training and inference. The review concludes with a discussion on specific applications and future research directions, emphasizing the versatility and potential of Gaussian Splatting in addressing various challenges in 3D reconstruction and NVS. The authors aim to provide a thorough overview of the latest advancements in Gaussian Splatting, highlighting its significance in the rapidly evolving field of 3D computer vision.