scipion-em-sphire 3.0.3

Cryolo picker, training and custom model importer. Janni denoising.

Sphire Scipion plugin

This plugin allows to use cryolo within Scipion framework. So far we have implemented:

• crYOLO particle picker

• JANNI denoising

This plugin is able to install cryolo (which includes janni) and the generic models for both picking and denoising.

crYOLO is a pipeline for particle detection in cryo-electron microscopy images which is based on the deep learning object detection system “You Only Look Once” (YOLO).

JANNI (Just Another Noise 2 Noise Implementation) implements a neural network denoising tool based on deep learning.

Setup

It is assumed that Scipion3 is currently installed. If not, follow the instructions here.

Plugin scipion-em-sphire requires to have Conda (Anaconda or Miniconda) installed and not initialized in the shell.

• Install this plugin in user mode:

  It can be installed in user mode via Scipion main window (Configuration > Plugins) or using the command line:

scipion installp -p local/path/to/scipion-em-sphire

• Install this plugin in developer mode:

scipion installp -p local/path/to/scipion-em-sphire --devel

Plugin integration

The following steps presuppose that you have Anaconda or Miniconda installed on your computer. In ~/.config/scipion/scipion.conf (Option View > Show Hidden Files must be enabled) set CONDA_ACTIVATION_CMD variable in the Packages section.

For example:

CONDA_ACTIVATION_CMD = . ~/anaconda2/etc/profile.d/conda.sh

If you wish to install the plugin with the default settings just go to plugin manager and install scipion-em-sphire. This will create the default environment named cryolo-x.x.x (where x.x.x is referred to the downloaded version of cryolo) for you.

You are ready to use crYOLO and JANNI. If you wish to change the environment name you can introduce CRYOLO_ENV_ACTIVATION variable in the ~/.config/scipion.conf variables section:

For example:

CRYOLO_ENV_ACTIVATION = conda activate cryoloenvname

Downloaded crYOLO and JANNI general models can be found, respectively, in the following locations:

<SCIPION_HOME>/software/em/cryolo_model-[model_version]

<SCIPION_HOME>/software/em/cryolo_negstain_model-[model_version]

<SCIPION_HOME>/software/em/janni_model-[model_version]

Running only with CPU

CrYOLO is able to run in CPU, however this is only recomended for predicting task, not trainning. For that reason, the CPU implementation is only enabled for the crYOLO-Picking protocol. In this protocol, both implementation are possible and you must select one of them in the GPU section of the from.

The CPU implementation of crYOLO is not installed by default. Therefore, to be able to run the CPU version of crYOLO-picking, you must install the cryoloCPU-[version] package in the Plugin Manager >> scipion-em-sphire or by running

scipion3 installb cryoloCPU

The CPU integration of crYOLO is installed under a conda environment called cryoloCPU-[version] . If you want to modify this environ name, please set the following variable in the scipion’s config file.

CRYOLO_ENV_ACTIVATION_CPU = conda activate cryoloCPU-envName

Running plugin tests

To check that everything is properly installed and configured, you might want to run some tests:

scipion test --grep sphire --run