A ray traced deformed bunny, rendered with 1024x1024 pixels and 100 rays per pixel in 6 hours, using a dense grid SDF representation.
The bunny rendered using the exact learned SDF as the distance function.
The bunny rendered using a sparse octree with trilinear interpolation between SDF sample points.
The bunny rendered using a dense grid with trilinear interpolation between SDF sample points.
The bunny rendered using as voxels of an octree.
The bunny rendered as a depth map.
The bunny rendered as a normal map.
The first few learned level sets of the deformed bunny SDF.
The bunny continuously deforming between three training poses by linearly interpolating the input deformation code.
A learned dragon in an undeformed pose.
A learned dragon in a second pose, deformed to a training pose using the training pose's deformation code.
A learned dragon in a third pose, deformed to a training pose using the training pose's deformation code.
A bar deforming between multiple training poses by interpolating the deformation code. The training deformations are as seen below.