Requirements

Requirements#

To use deepRetinotopy, ensure the following dependencies are available:

Installation#

In general, there are two distinct ways to install and use deepRetinotopy: either through virtualization/container technology, that is Docker or Singularity, or via Neurodesk. Once you are ready to run deepRetinotopy, see Usage for details.

Docker#

If you prefer running deepRetinotopy locally via Docker, you can pull our container from Dockerhub and run it using the following commands:

docker pull vnmd/deepretinotopy_1.0.18
docker run -it -v ~:/tmp/ --name deepret -u $(id -u):$(id -g) vnmd/deepretinotopy_1.0.18

If you would like Python scripts to print output to the terminal in real-time, you can set the appropriate environment variable when running the container:

docker run -e PYTHONUNBUFFERED=1 -it -v ~:/tmp/ --name deepret -u $(id -u):$(id -g) vnmd/deepretinotopy_1.0.18

Once in the container (the working directory is deepRetinotopy_TheToolbox), you can run deepRetinotopy:

deepRetinotopy -s $path_freesurfer_dir -t $path_hcp_template_surfaces -d $dataset_name -m $maps

Singularity#

Alternatively, you can also download the Singularity/Apptainer container using the following command to run it locally or on your HPC:

date_tag=20250902
export container=deepretinotopy_1.0.18_$date_tag
curl -X GET https://neurocontainers.neurodesk.org/${container}.simg -O

Then, you can execute the container (so long as Singularity/Apptainer is already available on your computing environment) using the following command:

apptainer exec ./deepretinotopy_1.0.18_$date_tag.simg deepRetinotopy -s $path_freesurfer_dir -t $path_hcp_template_surfaces -d $dataset_name -m $maps

GPU inference#

You can use the same container for CPU-based inference pipeline as well as GPU-based inference.

To run our tool using a GPU, you need to pass the --nv flag:

apptainer exec --nv ./deepretinotopy_1.0.18_$date_tag.simg deepRetinotopy -s $path_freesurfer_dir -t $path_hcp_template_surfaces -d $dataset_name -m $maps

Neurodesk#

You can run deepRetinotopy on Neurodesktop or using Neurocommand through the following commands:

ml deepretinotopy/1.0.18
deepRetinotopy -s $path_freesurfer_dir -t $path_hcp_template_surfaces -d $dataset_name -m $maps

This method allows you to leverage the pre-configured environment provided by Neurodesk, ensuring compatibility and ease of use.

Contents