Abstract
We present BaRe-ESA, a novel Riemannian framework for human body scanrepresentation, interpolation and extrapolation. BaRe-ESA operates directly onunregistered meshes, i.e., without the need to establish prior point to pointcorrespondences or to assume a consistent mesh structure. Our method relies ona latent space representation, which is equipped with a Riemannian(non-Euclidean) metric associated to an invariant higher-order metric on thespace of surfaces. Experimental results on the FAUST and DFAUST datasets showthat BaRe-ESA brings significant improvements with respect to previoussolutions in terms of shape registration, interpolation and extrapolation. Theefficiency and strength of our model is further demonstrated in applicationssuch as motion transfer and random generation of body shape and pose.