Learning Deformable Tetrahedral Meshes for 3D Reconstruction

Neural Information Processing Systems 

Previous work on neural 3D reconstruction demonstrated benefits, but also limitations, of point cloud, voxel, surface mesh, and implicit function representations.