Saving and exporting models

CAREamics saves model checkpoints automatically during training. These checkpoints can be reused to resume prediction, or the model can be exported to the BioImage Model Zoo (BMZ) format to be shared with the community.

Checkpoints

During training, CAREamics uses a ModelCheckpoint callback that periodically saves the model weights to disk. Checkpoints are stored under the work_dir passed to the CAREamist as follows:

<work_dir>/checkpoints/<experiment_name>_<run_version>/
├── <experiment_name>_<epoch>_step_<step>_<val_loss>.ckpt
└── <experiment_name>_last.ckpt

The intermediate checkpoints correspond to the epochs with the lowest validation loss, while <experiment_name>_last.ckpt is the checkpoint from the final epoch.

Listing checkpoints

The get_checkpoints method returns the available checkpoint paths, sorted by epoch number, with the last checkpoint (if present) appended at the end.

checkpoints = careamist.get_checkpoints()

Reloading a checkpoint

A CAREamist can be recreated from any checkpoint path, for instance to run prediction in a new session.

best_careamist = CAREamist(checkpoint_path=checkpoints[0])

See Predicting with CAREamics for how to select the best or last checkpoint at prediction time.

Exporting to the BioImage Model Zoo

The export_to_bmz method packages the trained model into a .zip archive that can be uploaded to the BioImage Model Zoo, making it usable from other tools such as Fiji, Ilastik and deepImageJ.

1. The input array must have the same axes as recorded in the configuration.

For a complete walkthrough of every argument, the archive contents, model validation and loading a model back, see the Exporting to the BioImage Model Zoo tutorial.