Installation and Download

System requirements

At our institute in Dortmund, crYOLO is running on the following operation systems:

Ubuntu 18.04 LTS
Ubuntu 20.04
CentOS 7

We don’t test it but it should run on Windows as well.

Moreover the following GPUs are used:

NVIDIA Titan V
NVIDIA GTX 1080
NVIDIA GTX 1080Ti
NVIDIA RTX 2080 TI
NVIDIA GV 100

As the GPU accelerated version of tensorflow does not support MacOS, crYOLO does not support it either.

crYOLO depends on CUDA Toolkit and the cuDNN library. These will be automatically installed during crYOLO installation.

Install crYOLO

Note

Please read the Complimentary Science Software License before using crYOLO. If you are interested in using crYOLO in a commercial context please contact stefan.raunser@mpi-dortmund.mpg.de

The following instructions assume that pip and anaconda or miniconda are available. In case you have a old cryolo environment installed, first remove the old one with:

conda env remove --name cryolo

The installation happens in three steps:

First your need to install crYOLO either for CUDA11 (step 1A) or CUDA 10 (step 1B)
Install napari and the boxmanager plugin
Link the napari installation into your crYOLO environment for convenience.

If you want to install the boxmanager locally on your computer without crYOLO, you can skip 1 and step 3.

1A. With CUDA 11

The official tensorflow 1.15x does not support CUDA 11. To get support for it, we need use a custom tensorflow version from nvidia. The following steps do explain how setup crYOLO with this custom nvidia.

The first step is to create a new virtual environment:

conda create -n cryolo -c conda-forge -c anaconda pyqt=5 python=3 numpy==1.18.5 libtiff wxPython=4.1.1  adwaita-icon-theme 'setuptools<66'

Activate the environment:

conda activate cryolo

Next you need to install a package that allows the installation of a custom tensorflow version from nvidia:

pip install nvidia-pyindex

To install crYOLO with CUDA 11 support you need to run:

pip install 'cryolo[c11]'

1B. With CUDA 10

Warning

Recent NVIDIA graphic cards (e.g. RTX30XX, A5000) do not longer support CUDA 10. Only use the CUDA 10 setup if your have specific reason for it.

The first step is to create a new virtual environment:

conda create -n cryolo -c conda-forge -c anaconda pyqt=5 python=3.7 cudatoolkit=10.0.130 cudnn=7.6.5 numpy==1.18.5 libtiff wxPython=4.1.1  adwaita-icon-theme

Activate the environment:

conda activate cryolo

In case you run crYOLO on a GPU run:

pip install 'cryolo[gpu]'

But if you want to run crYOLO on a CPU run:

pip install 'cryolo[cpu]'

Warning

In case you run into glibc errors, you can find a solution in our troubleshooting section

Hint

You can also integrate crYOLO as Environment Module

2. Install napari and the boxmanager plugin

This step creates an environment for napari and installs the boxmanger plugin into it. If you already have a napari environment with the latest version (>=0.4.17), you can also use this and just do the last step of the following commands:

conda create -y -n napari-cryolo -c conda-forge python=3.10 napari=0.4.17 pyqt pip
conda activate napari-cryolo
pip install napari-boxmanager

Hint

napari_boxmanager will not work via x-forwarding

You may be used to using the cryolo_boxmanager via x-forwarding (by connecting via ssh -X / ssh -Y to your server). However, x-forwarding does not support OpenGL and therefore it does not work with the napari_boxmanager. In those cases, you need a local installation of the naparai boxmanager. To do that, just repeat this step on your local PC.

3. Link napari

This is an optional step, but for convenience reasons we link an adapted napari call into the crYOLO environment. With that you don’t need to switch environments when working with crYOLO. While this is optional, I assume during the tutorials that you did this step. Here is what you need to do:

conda activate cryolo
cryolo_dir=$(realpath $(dirname $(which cryolo_predict.py)))
napari_link_file=${cryolo_dir}/napari_boxmanager
conda activate napari-cryolo
echo -e "#\!/usr/bin/bash\nexport NAPARI_EXE=$(which napari)\nnapari_exe='$(which napari_boxmanager)'\n\${napari_exe} \"\${@}\""> ${napari_link_file}
chmod +x ${napari_link_file}
conda activate cryolo

That’s it!

You might want to check if everything is running as expected. Here is a reference example:

Reference example with TcdA1

Download the general models

We provide three general models. One for cryo-EM images which was trained on low-pass filtered images, another one for cryo-EM images but trained for images denoised by JANNI and one for negative stain images.

For cryo images (low-pass filtered)

Datasets: 43 real, 10 simulated, 10 particle free datasets on various grids with contamination
Uploaded: 27 May 2020
Download: ftp https
Config: Commands to create the config file can be found here.

For cryo images (neural network denoised with JANNI)

Datasets: 43 real, 10 simulated, 10 particle free data sets on various grids with contamination
Uploaded: 27 May 2020
Download: ftp https
Config: Commands to create the config file can be found here.

For negative stain images

Datasets: 10 real data sets
Uploaded: 26 February 2019
Download: ftp https
Config: Commands to create the config file can be found here.