CARE_on_image.py

Help Menu

usage: CARE_on_image.py [-h] [--matlab] [--matlab_pid MATLAB_PID] --data-path DATA_PATH [--overwrite]
                      --model-name {CARE-1Channel} [--model-version MODEL_VERSION]
                      --mlflow-tracking-direc MLFLOW_TRACKING_DIREC
                      [--mlflow-tracking-ip MLFLOW_TRACKING_IP]
                      [--mlflow-tracking-port MLFLOW_TRACKING_PORT]

Use CARE (content-aware image restoration) to denoise data. Data can be a single tiff or a folder of
tiffs. Uses MLFlow (v. 2-22-2) Registered Models to track and retrieve models.

options:
  -h, --help            show this help message and exit

Meta Parameters:
  Parameters determining how the Python inference session is run.

  --matlab              Whether to coordinate with a running MATLAB/FLAME `sessionPostProcessing` thread.
  --matlab_pid MATLAB_PID
                        The MATLAB process id to be used during engine linkage. Only required if '--
                        matlab' requested.
  --data-path DATA_PATH
                        The path to the data to infer on if '--matlab' is not requested.
  --overwrite           whether to overwrite the provided tiffs or save them in a separate folder at the
                        same level in the file directory.

Model Information:
  Variables to configure the name and version of the requested CARE model.

  --model-name {CARE-1Channel}
                        The name of the 'Registered Model' to pull from.
  --model-version MODEL_VERSION
                        The version of the registered model to pull from. If none is provided, most recent
                        is used.

MLFlow Tracking:
  Variables to configure MLFlow tracking for model retrieval.

  --mlflow-tracking-direc MLFLOW_TRACKING_DIREC
                        Directory with 'mlruns' folder. Try
                        '<mount>/SynologyDrive/CARE_for_MATLAB/mlruns'.
  --mlflow-tracking-ip MLFLOW_TRACKING_IP
                        IP address where to host the MLFlow server.
  --mlflow-tracking-port MLFLOW_TRACKING_PORT
                        Port where to run the MLFlow server.

Examples

1. Using the most recent CARE-1Channel model (registered in MLFlow tracking server at mlflow-tracking-direc) to infer on a folder of images.

python CARE_on_image.py --data-path path/to/data/directory --model-name CARE-1Channel --mlflow-tracking-direc path/to/tracking/directory

2. Inferring on a single image using a specific version of CARE-1Channel model registered in an MLFlow tracking server hosted on a specified ip and port.

python CARE_on_image.py --data-path path/to/image.tiff --model-name CARE-1Channel --model-version 2 --mlflow-tracking-direc path/to/tracking/directory --mlflow-tracking-ip 146.12.67.1 --mlflow-tracking-port 6700