CebraEM

Installation

Support for Linux and Windows

Note that CebraEM is developed on Linux. We have tested CebraEM on Windows and are currently working on a stable solution for Windows machines.

Note that for Windows 10, a Windows Subsystem for Linux (WSL) has to be installed for CebraEM to run. In Windows 11 this is already installed by default.

Installation of the MoBIE browser

Follow the instruction here to install the Fiji and the MoBIE browser which includes these steps:

• Download Fiji for your operating system
• Unpack Fiji to any location on the filesystem
• Open Fiji and install MoBIE:
  • Open the Fiji upater using Help -> Update... and wait for the updater to start up (may take a moment)
  • An information prompt will probably report that 'Your ImageJ is up to date!'. Press OK
  • In the "ImageJ Updater" window click Manage Update Sites
  • In the "Manage Update Sites" window scroll down or search for "MoBIE" and tick the box next to it
  • Press Apply and Close
  • Back in the "ImageJ Updater" window click Apply Changes which will now trigger download and installation of the required packages
  • Close and re-open Fiji to complete the installation

Alternatively, can download and extract a Fiji that already includes MoBIE from the this release.

After successful installation of MoBIE you can open a CebraEM/MoBIE project using Fiji -> Plugins -> MoBIE -> Open -> Project -> Open MoBIE Project ...

Download of the CebraEM source code

Here, you have two options:

1. Download the source code in the latest release

• Now unpack the contents to some path of your choice (here I'm now assuming you created a folder called src in the user directory)
• The path to CebraEM will now look something like this: ~/src/CebraEM-v0.0.1/ (Linux) or C:\Users\username\CebraEM-v0.0.1\ (Windows)
• Remember this path, this will be referred to below as /path/to/CebraEM

OR

2. Clone the CebraEM repo into a path of your choice:

cd /path/to/  # !!Replace the folder name!!
git clone https://github.com/jhennies/CebraEM.git

The path to CebraEM will be referred to below as /path/to/CebraEM

Installation of Miniconda

Install miniconda on your computer.

I generally recommend using the mamba package for installing the conda environments. Open a terminal and install mamba into the base environment with:

conda install -c conda-forge mamba

For the following descriptions an installation of Mamba is assumed (otherwise just replace "mamba" -> "conda")

CebraEM conda environment

These commands install CebraEM as well as all dependencies except pytorch (run each of these lines in your terminal/prompt):

cd /path/to/CebraEM
mamba env create -f environment.yaml
conda activate cebra-em-env

pip install -e ./cebra-em-core/
pip install -e ./cebra-ann/
pip install -e ./cebra-em/

For pytorch you will need to determine your Cuda version and select the correct installation call. Please refer to https://pytorch.org/get-started/locally/ for further information. As an example, for our system with Cuda version 11, we installed pytorch successfully using this pip install command:

pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

Now it is important to test the installation

The following script runs the main steps of the CebraEM pipeline using a small test dataset.

python test_installation.py

Possible errors:

ModuleNotFoundError: No module named 'torch'

The installation of pytorch is missing or did not work properly. Run nvidia-smi to determine your Cuda version and refer to https://pytorch.org/get-started/locally/ to determine the proper pytorch installation commands

CebraEM

CebraEM readme

Contains:

• Usage of CebraEM

CebraANN

CebraANN readme

Contains:

• Specific installation for the cebra-ann package in case CebraANN needs to be run on a separate machine
• Usage of CebraANN

CebraNET

CebraNET readme