Skip to content

copick convert nnunet

torch

Convert a copick project to nnUNet raw dataset format.

Plugin command — copick-torch

This command is provided by the copick-torch plugin, not copick core. Install it to make this command available:

pip install copick-torch

See the plugin system guide for details.

Usage

copick convert nnunet [OPTIONS]

Description

Reads tomograms and segmentation masks from a copick project and writes them as .nii.gz files in the nnUNet raw dataset layout (a Dataset{id}_{name} folder containing imagesTr, labelsTr, and optionally imagesTs). A dataset.json describing the channel and label map is generated from the targets config stored in the copick overlay.

Training runs default to every run not listed in the test set. The tomogram voxel spacing is read from the tomogram URI and converted from Angstroms to nanometres so that nnUNet's patch-size planner sees reasonable numbers.

Options

Option Type Default Description
-c, --config path required Path to copick config.json
-uri, --tomo-uri text wbp@10.0 Tomogram URI to use for training
-sinfo, --seg-info text targets Segmentation info as 'name' or 'name,user_id,session_id'
-train, --train-run-ids text — Training run IDs, e.g. run1,run2,run3. Default: all runs not in test set.
-test, --test-run-ids text — Test run IDs, e.g. run4,run5
-id, --dataset-id integer 1 nnUNet dataset ID (integer; becomes Dataset{id}_{name})
-n, --dataset-name text required nnUNet dataset name
-o, --output path required Path to nnunet_raw output directory
-j, --num-workers integer 4 Number of parallel worker threads for converting tomograms.

Examples

# Convert all runs in a project to an nnUNet dataset
copick convert nnunet -c config.json --dataset-name Membrane \
    --output /data/nnunet_raw

# Convert with an explicit tomogram URI, segmentation info, and dataset id
copick convert nnunet -c config.json --tomo-uri wbp@10.0 \
    --seg-info targets,nnunet,1 --dataset-id 5 --dataset-name Membrane \
    --output /data/nnunet_raw

# Convert with an explicit train/test split and more worker threads
copick convert nnunet -c config.json --dataset-name Membrane \
    --train-run-ids run1,run2,run3 --test-run-ids run4,run5 \
    --output /data/nnunet_raw -j 8

See also