Inferring Neural Signed Distance Functions by Overfitting on Single Noisy Point Clouds through Finetuning Data-Driven based Priors

It is an important task to estimate an implicit function from a point cloud in computer graphics, computer vision, and robotics. An implicit function, such as a signed distance function (SDF), describes a continuous 3D distance field to indicate distances to the nearest surfaces at arbitrary locations. Since point clouds are easy to obtain, they are widely used as an information source to estimate SDFs, particularly without using normals that are not available for most scenarios.