3D-CODED: 3D Correspondences by Deep Deformation

Abstract

We present a new deep learning approach for matching deformable shapes by introducing Shape Deformation Networks which jointly encode 3D shapes and correspondences. This is achieved by factoring the surface representation into (i) a template, that parameterizes the surface, and (ii) a learnt global feature vector that parameterizes the transformation of the template into the input surface. By predicting this feature for a new shape, we implicitly predict correspondences between this shape and the template. We show that these correspondences can be improved by an additional step which improves the shape feature by minimizing the Chamfer distance between the input and transformed template. We demonstrate that our simple approach improves on stateof-the-art results on the difficult FAUST-inter challenge, with an average correspondence error of 2.88cm. We show, on the TOSCA dataset, that our method is robust to many types of perturbations, and generalizes to non-human shapes. This robustness allows it to perform well on real unclean, meshes from the the SCAPE dataset.

Efficient Learning on Point Clouds With Basis Point Sets

“…3). These numbers are comparable, but higher than state-of-the-art methods like [15] or [49]. However, we note that the two methods outperforming BPS in the FAUST intra challenge are orders of magnitude slower than our system.…”

“…However, they are typically computationally expensive and require the use of a deformable model at application time. Machine learning based methods [15] remove this dependency by replacing them with a sufficiently large training corpus. However, current solutions like [15] rely on multistage models with complex internal representations, which makes them slow to train and test.…”

“…Machine learning based methods [15] remove this dependency by replacing them with a sufficiently large training corpus. However, current solutions like [15] rely on multistage models with complex internal representations, which makes them slow to train and test. We encourage the reader interested in human mesh registration to review the excellent summary of previous work provided in [15].…”

Geometric Disentanglement for Generative Latent Shape Models

“…We generate 3200 meshes for each of the five categories by sampling from each cluster distribution. Following the generation procedure in other works [21], we then sample the pose (here, the joint angles) via a Gaussian distribution with a standard deviation of 0.2. We then split the resulting dataset into 15000 training and 1000 testing meshes (each comprised of equal numbers of meshes per species).…”

“…the protocol from 3D-CODED [21]. Briefly, we sampled 20500 meshes from each of the male and female models using random samples from the SURREAL dataset [60].…”