Supplementary Materials for: NCP: Neural Correspondence Prior for Effective Unsupervised Shape Matching

Our work uses the functional map framework as a first estimator for p2p maps, and multiple losses on point-to-point (p2p) maps to learn robust features that allow extracting good correspondences using the nearest neighbor in feature space. We provide a brief overview in the next section. Functional maps The functional map (fmap) framework was used for the first stage of our NCP-UN algorithm.