3D Gaussian Splatting is an innovative rasterization technique that revolutionizes the way we render photorealistic scenes in real-time. Unlike traditional methods that rely on drawing numerous triangles to depict a scene, this technique uses gaussians—a mathematical concept often used in statistics to represent data distributions. In simple terms, imagine replacing the tiny triangles typically used in graphics with small, adjustable blobs (gaussians) that can vary in size, shape, color, and transparency. These blobs are strategically placed and manipulated to construct an incredibly detailed and realistic 3D scene from a relatively small set of image samples. The process begins by converting a collection of 2D images into a 3D point cloud, which is then transformed into these gaussian blobs. Through a training phase, similar to methods used in machine learning, the characteristics of each gaussian are finely tuned to match the target scene as closely as possible. This method not only offers a new level of detail and realism in real-time rendering but also opens up exciting possibilities for future graphics technologies. Its potential applications extend beyond traditional graphics, potentially influencing fields like Embodied AI, where understanding and representing 3D spaces is crucial.