Installation

Photo by Brett Jordan on Unsplash

To install VTK, you need to have a few things prepared in advance. Mainly, you need Nvidia CUDA installed first; check the README file to see how to install it.

OpenCV and TensorFlow are the essential dependencies of VTK. You have a few options here:

• Build both OpenCV and TensorFlow from source for maximum optimization (recommended for those with nothing else to do with their time).
• Build just TensorFlow from source for maximum speed on inference (recommended for those willing to do so).
• Install prebuilt varieties of both TensorFlow and OpenCV from PyPI (recommended for beginners).

We will not be covering building OpenCV and TensorFlow from source, as every computer setup is different and every software difference contributes to a different solution being used.

To install the beginner’s set of dependencies with VTK, follow the instructions below:

1. Install Python 3.6 (or newer) on your machine. Do not use Python 2. It will not work and issues resulting from Python 2 will be closed as “won’t fix”.
2. Using the virtualenv tool, which you can install in a terminal using the command pip install virtualenv, create a new virtual environment:
   • virtualenv -p python3 venv
3. Activate this new virtual environment using the source command:
   • source venv/bin/activate
4. Clone VTK from GitHub using the git command:
   • git clone https://github.com/Robocubs/vtk
5. Enter the VTK folder and install all required packages using the pip3 command:
   • cd vtk
   • pip3 install -r requirements.txt
   • or pip3 install -r .travis-requirements.txt if your computer does not have an NVIDIA GPU
6. Verify that the requirements have been correctly installed by running VTK’s test suite:
   • cd tests
   • coverage run --source=../vtk/ -m nose2
   • cd ..
7. Finally, install VTK in your virtual environment using the below commands:
   • python3 setup.py install

Alright! Now, it’s time to start using VTK in your project.