3D Gaussian Splatting vs NeRF: The End Game of 3D Reconstruction?

3D Gaussian Splatting is a promising new technique for 3D reconstruction that may surpass previous methods like NeRFs. It initializes 3D Gaussians from a point cloud, projects them to 2D images, and optimizes to match real images. Key advantages include starting from an explicit 3D structure and allowing editing. While showing impressive results, including applications in self-driving cars, it's unclear if it's the ultimate solution for 3D reconstruction. The field continues to evolve rapidly.