Train SwinUnetr Pixel Embedding Network

Train the SwinUnetr pixel embedding network using the provided script and dataset.

Arguments

--lr

Learning rate. (default value: PARAMETER_VALUE)

--logdir_path

Path to save logs and checkpoints. (default value: PARAMETER_VALUE)

--batch_size

Batch size for training. (default value: PARAMETER_VALUE)

--patch_threshold

Patch threshold. (default value: PARAMETER_VALUE)

--loss_temperature

Loss temperature. (default value: PARAMETER_VALUE)

--pretrained_weights_path

Path to pretrained weights. (default value: PARAMETER_VALUE)

--max_epochs

Maximum number of epochs for training. (default value: PARAMETER_VALUE)

--copick_config_path

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

--train_run_names

Comma-separated list of Copick run names for training data. (default value: PARAMETER_VALUE)

--val_run_names

Comma-separated list of Copick run names for validation data. (default value: PARAMETER_VALUE)

--voxel_spacing

Voxel spacing to be used. (default value: PARAMETER_VALUE)

--tomo_type

Type of tomogram to process. (default value: PARAMETER_VALUE)

--session_id

Session ID for accessing Copick data. (default value: PARAMETER_VALUE)

--user_id

User ID for accessing Copick data. (default value: PARAMETER_VALUE)

--segmentation_name

Name of the segmentation to use from Copick. (default value: PARAMETER_VALUE)

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