Labkit Command Line

A command line version of Labkit, that allows to segment large images (up to several terrabytes) on a cluster.

Usage

It's simple to use. You need to have Java 8 and Snakemake installed on your computer. Then installing Labkit command line just means to download and extract this zip archive. It contains the Labkit command line binary and a complete example that uses Snakemake:

├── input
│   ├── image.xml               # Example Image stored as Big Data View XML + HDF5
│   ├── image.h5                
│   └── drosophila.classifier       # Example classifier trained using the Labkit FIJI Plugin.
├── labkit-command-line-0.1.1.jar   # Labkit Command Line binary
└── Snakefile                       # Snakemake file

To run the example: Open a terminal, go to the directory (containing the extracted files) and run:

$ snakemake

To segment your data, just replace the example image with your data. Check the settings in the "Snakefile" if it fits your data. Note: The "tmp" and "output" folder need to be deleted before you rerun snakemake with new data.

Usage on the Cluster

Requirements: Snakemake & Java 8 need to be installed on the cluster.

Just download and extract the zip file mentioned above to a directory that's shared between the cluster nodes. To run the exmple on a SLURM Cluster, use this command:

$ snakemake --cluster=sbatch --jobs=10 --local-cores=1 --restart-times=10

To run on a PBS Cluster, use:

$ snakemake --cluster=qsub --jobs=10 --local-cores=1 --restart-times=10

Prerequirements

Java 8

Labkit command line tool requires Java 8, it doesn't work with any other Java version.

Snakemake

It's recommended to use the Labkit command line tool together with Snakemake. If you are running Ubuntu, Snakemake can be installed using the usual package manager:

$ sudo apt-get install snakemake

juglab / labkit-command-line