Skip to main content

MONAILabel Server Installation

To install MONAI Label along with additional models for higher-level morphological structure segmentation, follow below steps.

Using Docker​

Docker prerequisites​

To deploy server using Docker you need to install Docker on your computer. This step is different for different operating systems. For this purpose follow the instructions in official Docker Documentation.

  1. Visit the Docker Documentation page and select the OS you have installed.

    Docker - select OS

  2. Follow the instructions on the appropriate page for the selected OS and install Docker.

Get Docker Image​

  1. Pull image from Docker Hub.
docker pull xvykopal/monailabel-server:latest

Without Docker​

General prerequisites​

These prerequisites are necessary when you want to run the server without Docker.

Python​

First of all, you need to install Python with the verion >=3.9 and setup the Python in your computer's settings.

  1. Install Python version >= 3.9.

  2. Add the path to Python to the environment variables. The path to Python added to the environment variables should look similar to the following: C:\Users\{user}\AppData\Local\Programs\Python\{python}, where {user} is replaced with the user identifier within the OS and {python} represents the directory name with the installed Python for a specific version.

    2.1. On Windows, open the Start Menu and type, e.g environment

    Set environment variables 1

    2.2 Select Edit the system envirnoment variables

    Set environment variables 2

    2.3. Edit Path variable mainly for the user, but sometimes it is necessary to edit Path variable also in system variables.

    Set environment variables 3

  3. Python should be now available. You can check it in console by running the following command: python --version, if there is no error it means that Python was successfully installed.

Source code​

The next step is to download source code from the GitHub repository.

  1. Download the development branch from the repository.

    Source code download

    • The first method via Code > Download ZIP

    Source code download

    • The second method via:

      git clone https://github.com/ivanvykopal/MONAILabel.git
      git fetch --all
      git checkout development

  2. After we downloaded the source code, we can create a new python environment:

    • Using the venv command:

      virtualenv ".venv/monailabel" -p python3.9.16

    • Using conda

      conda create -n "monailabel" python=3.9.16

      It is recommended to use Anaconda or Miniconda for managing multiple environments.

  3. Install the packages from requirements.txt

    • venv

      .venv/monailabel/Scripts/activate
      python -m pip install --upgrade pip
      pip install -r requirements.txt

    • conda

      conda activate monailabel
      python -m pip install --upgrade pip
      pip install -r requirements.txt

  4. To run server, it is needed to download openslide binaries and extract them to any location. After extraction, add the path to the bin directory to the environment variables in PATH, as you done for Python.

  5. To add the path to the monailabel\scripts directory to the PATH environment variable, follow the instructions in this macOS example. If any dll libraries are missing, you can download them from dll-files.com. Specifically, if you need the missing cudnn64_8.dll, you can download it using this link. After downloading the dll files, you need to copy them to the system disk. On Windows systems, the standard location is C:\Windows\System32.

Weights and Sample Project​

To run the inference of the added models, we need to download weigths of the models. To download them, it is necessary to contact Ivan Vykopal or VGG Research group

  1. Download the weights of the models and sample project provided by authors.

  2. Add the downloaded weights for pathology_structure_segmentation_nestedunet to apps/pathology/model/pathology_structure_segmentation_nestedunet/models and add the weights for pathology_structure_segmentation_deeplabv3plus to apps/pathology/model/pathology_structure_segmentation_deeplabv3plus/models.