Scripting in Python

TARDIS-em library can be use to simply and fast script your own workflows.

More examples can be find in: tardis_em/examples/TARDIS_em_Script.ipynb

Example

from tardis_em.utils.predictor import GeneralPredictor

predictor = GeneralPredictor(
    predict: str,
    dir_s: Union[str, tuple[np.ndarray], np.ndarray],
    binary_mask: bool,
    output_format: str,
    patch_size: int,
    convolution_nn: str,
    cnn_threshold: float,
    dist_threshold: float,
    points_in_patch: int,
    predict_with_rotation: bool,
    instances: bool,
    device_s: str,
    debug: bool,
    checkpoint: Optional[list] = None,
    correct_px: float = None,
    amira_prefix: str = None,
    filter_by_length: int = None,
    connect_splines: int = None,
    connect_cylinder: int = None,
    amira_compare_distance: int = None,
    amira_inter_probability: float = None,
    tardis_logo: bool = True,
)

semantic, instance, instance_filter = predictor()

predict: File directory to visualize.

• Allowed options: Microtubule, Membrane2D, Membrane

-dir_s: Directory to a single file, folder with files or numpy array with tomogram/micrograph.

• Allowed options: str, np.ndarray

-binary_mask: If True, Predictor assume, that input images are binary mask. The semantic segmentation step would be skipped and only instance segmentation results will be produce.

• Allowed options: bool

-output_format: Two output format for semantic and instance prediction.

• Tips: Define as semantic_instance format. For example to output semantic segmentation mask as .mrc file format, and instance segmentation as .csv file. Type mrc_csv

• Allowed options Semantics: None, am, mrc, tif, npy

• Allowed options Instances: None, am, mrc, tif, npy, amSG, csv, stl

-patch_size: Image crop size used during semantic segmentation.

• Allowed options: int

-convolution_nn: Type of pre-train CNN model.

• Allowed options: unet, fnet_attn

-cnn_threshold: Threshold for CNN model. Used during semantic segmentation.

• Allowed options: float

-dist_threshold: Threshold for DIST model. Used during instance segmentation.

• Allowed options: float

-points_in_patch: Maximum number of points per patched point cloud.

• Tip: About 1000 points require ~ 12Gb of GPU or RAM (if device == ‘cpu’)

• Allowed options: int

-predict_with_rotation: If True, CNN predict with 4 90* rotations.

• Allowed options: bool

-instances`: If True, run instance segmentation after semantic.

• Allowed options: bool

-device_s: Device on which prediction will take place.

• Allowed options: cpu, gpu or number between 0-9 indicating gpu id

-debug: If True, enable debugging mode which save all intermediate files.

• Allowed options: bool

-checkpoint: List of model checkpoints for semantic and instance segmentation. If its None, TARDIS retrieves weights from AWS.

• Default: None

• Allowed options: list[str], list[dict]

-correct_px: Indicate correct pixel size for image data. If its None, TARDIS retrieves pixels size from the file header.

• Default: None

• Allowed options: float, None

-amira_prefix`: Optional, Amira file prefix name used for spatial graph comparison.

• Default: None

• Allowed options: str, None

-filter_by_length: Optional, filter setting for filtering short splines. Value given in Angstrom.

• Default: None

• Allowed options: int, None

-connect_splines: Optional, filter setting for connecting near splines. Value given in Angstrom.

• Default: None

• Allowed options: int, None

-connect_cylinder: Optional, filter setting for connecting splines withing cylinder radius. Value given in Angstrom.

• Default: None

• Allowed options: int, None

-amira_compare_distance: Optional, compare setting, max distance between two splines to consider them as the same. Value given in Angstrom.

• Default: None

• Allowed options: int, None

-amira_inter_probability: Optional, compare setting, portability threshold to define comparison class. Value given between 0-1 as a probability.

• Default: None

• Allowed options: float, None

-tardis_logo: If True, GeneralPredictor will display terminal or command-line logs.

• Default: True

• Allowed options: bool