Train 3D ResUNet for Segmentation with Copick Dataset

Train a 3D ResUNet network using the Copick dataset for segmentation.

Arguments

--copick_config_path

Path to the Copick configuration file (default value: PARAMETER_VALUE)

--train_run_names

Names of the runs in the Copick project for training (default value: PARAMETER_VALUE)

--val_run_names

Names of the runs in the Copick project for validation (default value: PARAMETER_VALUE)

--tomo_type

Tomogram type in the Copick project (default value: PARAMETER_VALUE)

--user_id

User ID for the Copick project (default value: PARAMETER_VALUE)

--session_id

Session ID for the Copick project (default value: PARAMETER_VALUE)

--segmentation_type

Segmentation type in the Copick project (default value: PARAMETER_VALUE)

--voxel_spacing

Voxel spacing for the Copick project (default value: PARAMETER_VALUE)

--lr

Learning rate for the ResUNet training (default value: 0.0001)

--logdir

Output directory name in the current working directory. Default is checkpoints (default value: checkpoints)

--experiment_name

mlflow experiment name. Default is resunet_experiment (default value: resunet_experiment)

--max_epochs

Maximum number of epochs for training (default value: 10000)

--batch_size

Batch size for training and validation (default value: 1)

--num_res_units

Number of residual units in the UNet (default value: 2)

Please follow this link for details on how to install and run this solution.