Global approaches to Structure from Motion (SfM) seek to recover camera poses and 3D structure in multiview settings by simultaneously handling information from all input images. In this talk, I will present two approaches to global SfM. The first approach relies on a complete algebraic characterization of essential/fundamental matrices in multiview settings. We use this characterization to introduce novel manifold averaging algorithms that jointly solve for camera positions, orientations, and calibration parameters. The second approach uses projective factorization, which we implement using a deep network with a suitable equivariant architecture.