Get the docker image

Get the docker image of the workflow from Dockerhub:

docker pull {remote_image}

Or, alternatively, build workflow Docker image from source (GitHub repository) Inside repository folder:

docker build -t {local_image}

.

Prepare the data folder

Prepare an empty folder {DATA_PATH} with a subfolder /data and subfolders:

• {DATA_PATH}/data/in add input images to this folder^*
• {DATA_PATH}/data/out workflow results are exported to this folder
• {DATA_PATH}/data/gt leave empty

^* Images should be 8/16-bit TIFF (2D) or 8/16-bit single file OME-TIFF (C,Z,T). The string _lbl is forbidden in image name since it is used to identify ground truth annotation images.

Run the workflow

Prepare an empty folder {DATA_PATH} with a subfolder /data and subfolders:

docker run -v {DATA_PATH}/data:/data -it {image_name} {WORKFLOW_PARAMETERS} --infolder /data/in --gtfolder /data/gt --outfolder /data/out --local

This whole procedure is illustrated in the following Python Jupyter notebook: https://github.com/Neubias-WG5/biaflows_jupyter_local

Notes:
• --local (-l): do not download nor upload any content from / to BIAFLOWS. The images (input and ground truth) are read from specified folders. Metrics are optionally displayed to standard output.
For a more fine-grained control over BIAFLOWS interactions:
• --no_download (-nd): images and ground truth are not downloaded from BIAFLOWS
• --no_annotations_upload (-nau): annotations are not uploaded to BIAFLOWS
• --no_metrics_computation (-nmc): metrics are not computed
• --no_metrics_upload (-nmu): metrics are not uploaded to BIAFLOWS.