installation_windows.md

Installation Instruction for Windows

(tested on Windows 10 with anaconda)

Step 0: Install Microscoft Build Tools

Go to Step 1 if you already have these installed (it is very likely that you already have it, if you have done any python or C++ programming on your machine).

Download and install the Build Tools from Microsoft

Note 1: Depending on the time you access this page, you may be directed to the download page of "Build Tools for Visual Studio 2017" or "Build Tools for Visual Studio 2018", etc. In general, as long as the version is at least 2015, any subsequent version should work.

Note 2: When you install the Build Tool, you may be asked to choose what to install. Only the tool for Visual Studio C++ (e.g, pack name: "MSVC v142 - VC 2019 C++") is required for the Allen Cell Segmenter. After installing the Build Tool, make sure to reboot your machine.

Step 1: Install conda

Go to Step 2 if you have anaconda or miniconda installed

Go to Install conda on Windows, choose Anaconda Installer (for Python 3) and then follow the installation instructions.

Note: What is conda and anaconda, and why we need this? Because conda can effectively manage environment and package installation, setting up conda will make the following steps straightforward and help avoid future problems (conda itself is also very easy to set up).

Step 3: Verify requirement and prepare for installing segmenter

Step 3.1: Start conda

All commands below are typed into Anaconda Prompt window

Step 2.2: Test conda version

conda info

You may see somthing like

conda version : 4.6.11
python version : 3.7.3.final.0

conda version > 4.4 is preferred. To update conda, check out how to update your conda. python version >=3.6 is required.

Step 2.3: Test git

git --version

If you don't have git, follow Git for Windows to install. It is okay to just use default installation settings. You will need to restart Anaconda Prompt after installing Git.

Step 2.4: Test pip

pip show pip

A message will be printed out on your screen. If you see a warning, like a newer version is available, you can follow the instruction to upgrade you pip.

Step 2.5: Create a new empty conda environment, which we will name "segmentation" (You can certainly choose a different name.)

conda create -n segmentation python=3.6

Step 2.6: Activate your conda environment "segmentation"

conda activate segmentation

Step 2.7: Install nb_conda (for easy conda environment management in jupyter notebook)

conda install nb_conda

Step 3: Install segmenter

Step 3.1: Clone aics-segmentation repository from Github (suppose you want to save the folder under 'C:\Projects')

cd C:\Projects
git clone https://github.com/AllenInstitute/aics-segmentation.git

Step 3.2: install the packages

cd ~/Projects/aics-segmentation
pip install numpy
pip install itkwidgets==0.14.0
pip install -e .[all]

Note 1: Please note that for the users with both python 2 and python 3 installed, use pip3 instead of pip in the commands

Note 2: We use the packge itkwidgets for visualizaiotn within jupyter notebook. Currently, we find version 0.14.0 has slightly better performance in visualizing segmentation results. If you find this viwer keeps crashing in your browser, try pip uninstall itkwidgets and then pip install itkwidgets==0.12.2. For JupyterLab users, version >= 0.17.1 is needed.

Note 3: For Jupyter Lab users, the itk viewer requires additionally run:

jupyter labextension install @jupyter-widgets/jupyterlab-manager itk-jupyter-widgets

Note 4: For advanced user to deploy segmenter on cluster, our package is also available on PyPi

Step 3.3: Test segmenter

cd C:\Projects\aics-segmentation\lookup_table_demo
jupyter notebook, choose anaconda installer.

This will take you to your default browser (e.g., Chrome) and launch Jupyter Notebook App within your browser. Open "test_viewer.ipynb" and test if you can run the notebook from beginning to the end. See more details on How to use Jupyter Notebook to running the workflow in the Look-up Table