This library is under construction.

The implicit lab is a python library for the computation of implicit neural representations (INR) of geometrical objects. It is built on top of pytorch and provides utilities for geometric dataset handling and sampling, neural architecture definitions, training, geometrical request and visualization.

Implicit neural representations are a recent technique for encoding a signal in the parameters of a neural network. This library focuses of neural representations of surfaces, which are defined as the zero level set of some continuous function over space, like the signed distance function.

The (WIP) documentation can be found here: https://GCoiffier.github.io/neural_implicit_lab/

Features

The library currently features:

• Loading, processing and sampling of various input geometry types (point clouds, polylines, surface meshes) to generate relevant training datasets
• Training of various neural models according to published algorithms
• Geometrical queries on INRs

Installation

To install implicitlab, simply type:

pip install implicitlab

Dependencies

Implicitlab currently depends on the following libraries:

• pytorch
• deel-torchlip
• meshio
• mouette
• libigl
• triangle
• numpy
• scipy
• scikit-image
• matplotlib