Usage

Install

To use fit-a-nef, you need to install it. To install it, you need to run the following command in a terminal:

(.venv) $ pip install .

It is important to note that jax and torch are dependencies to this library, however, they might be installed with the correct version. If this happens, try installing jax and torch first, and only then run the command above.

Installing the repository

After the installation of the library, you can install the dependencies to run the code in the repository.

The main additional dependencies are wandb, optuna, and ml_collections. To install them, run the following command:

(.venv) $ pip install wandb optuna ml_collections

The, for the shape dataset we use trimesh to render the shapes, and a marching cubes algorithm based on it that is available in the dataset/shape_dataset/utils folder. To install them, run the following command:

(.venv) $ pip install trimesh
(.venv) $ cd dataset/shape_dataset/utils
(.venv) $ python setup.py build_ext --inplace

If problems arise when installing or running the marching cube algorithm, it could be due to the CUDA version being used. In that case, consider simply removing this plotting from the code. It is possible to visualize shapes by plotting the point cloud, marching cubes are not necessary.

Now you can look at :ref:fitting_neural_fields for more information on how to actually train your first models.