Segmentation

Training from Scratch

Training a segmentation model with the default configuration:

CLI

asp_train_seg task=Task004_Name \
    +model=unet_b \
    data.train_split=split_75_15_10

With a custom config file

asp_train_seg --config-name my_seg_config

Restarting a failed run

Find the run_id from the output directory or Hydra logs, then rerun with run_id=<id>:

asp_train_seg task=Task004_Name \
    +model=unet_b \
    data.train_split=split_75_15_10 \
    run_id=123

Key training parameters

Parameter	Description
`task`	Task name (folder in `$ASPARAGUS_DATA`)
`+model`	Model architecture config
`data.train_split`	Split file name
`training.patch_size`	3D patch size for patch-based training
`training.epochs`	Total number of training epochs
`training.seed`	Random seed for reproducibility
`run_id`	Assign a specific run ID (optional)

Finetuning from a Pretrained Model

Fine-tuning initialises the model encoder (and optionally the decoder) from a pretrained checkpoint.

CLI (by run_id)

asp_finetune_seg \
    task=Task004_Name \
    checkpoint_run_id=435850 \
    load_checkpoint_name=last.ckpt

CLI (by path)

asp_finetune_seg \
    task=Task004_Name \
    checkpoint_path=/path/to/model.ckpt

Finetuning-specific parameters

Parameter	Description
`checkpoint_run_id`	Run ID of the pretrained checkpoint
`checkpoint_path`	Absolute path to checkpoint (alternative to `run_id`)
`load_checkpoint_name`	Checkpoint filename (e.g., `last.ckpt`, `best.ckpt`)
`training.repeat_stem_weights`	Repeat stem weights to adapt 2D→3D or channel mismatch

Note

Fine-tuning uses a separate, typically lower learning rate (model.finetune_lr) and optimizer (model.finetune_optim) compared to training from scratch.

Testing / Inference

Run inference on a held-out test set using a trained or fine-tuned checkpoint:

CLI

asp_test_seg \
    test_task=Task004_Name \
    checkpoint_run_id=1234 \
    load_checkpoint_name=last.ckpt \
    test_split=TEST_75_15_10

Switching checkpoints

To test with a specific checkpoint epoch:

asp_test_seg \
    test_task=Task004_Name \
    checkpoint_run_id=1234 \
    load_checkpoint_name=epoch=4-step=25.ckpt \
    test_split=TEST_75_15_10

Outputs

Predictions are saved to $ASPARAGUS_MODELS/<run_id>/predictions/ as NIfTI files, mapped back to the original image coordinate space via reverse preprocessing.

Evaluation at Scale

For evaluating across multiple tasks and checkpoints, use the EvalBox:

The EvalBox →