Deep Regression on Manifolds: A 3D Rotation Case Study

Romain Brégier

Many machine learning problems involve regressing variables on a non-Euclidean manifold -- e.g. a discrete probability distribution, or the 6D pose of an object. One way to tackle…