Deep Meta Functionals for Shape Representation
This addresses the problem of accurate 3D shape representation from limited 2D data for computer vision and graphics applications, offering a novel approach but with incremental gains in accuracy.
The paper tackles 3D shape reconstruction from a single image by introducing a deep neural network that maps an image to network weights, which then classify points in a volume to represent shapes with unlimited capacity and resolution, resulting in more accurate inference than existing methods like voxel-, silhouette-, and mesh-based approaches.
We present a new method for 3D shape reconstruction from a single image, in which a deep neural network directly maps an image to a vector of network weights. The network \textcolor{black}{parametrized by} these weights represents a 3D shape by classifying every point in the volume as either within or outside the shape. The new representation has virtually unlimited capacity and resolution, and can have an arbitrary topology. Our experiments show that it leads to more accurate shape inference from a 2D projection than the existing methods, including voxel-, silhouette-, and mesh-based methods. The code is available at: https://github.com/gidilittwin/Deep-Meta