deepblink.inference module

Model prediction / inference functions.

deepblink.inference.get_intensities(image: numpy.ndarray, coordinate_list: numpy.ndarray, radius: int, method: str = 'sum') → numpy.ndarray

Finds integrated intensities in a radius around each coordinate.

Parameters:

• image – Input image with pixel values.
• coordinate_list – List of r, c coordinates in shape (n, 2).
• radius – Radius of kernel to determine intensities.
• method – How the integrated intensity should be calculated [options: sum, mean, std].

Returns:

Array with all integrated intensities.

deepblink.inference.get_probabilities(matrix: numpy.ndarray, coordinates: numpy.ndarray, image_size: int = 512) → numpy.ndarray

Find prediction probability given the matrix and coordinates.

Parameters:

• matrix – Matrix representation of spot coordinates.
• coordinates – Coordinates at which the probability should be determined.
• image_size – Default image size the grid was layed on.

Returns:

Array with all probabilities matching the coordinates.

deepblink.inference.predict(image: numpy.ndarray, model: keras.engine.training.Model, probability: Union[None, float] = None) → numpy.ndarray

Returns a binary or categorical model based prediction of an image.

Parameters:

• image – Image to be predicted.
• model – Model used to predict the image.
• probability – Cutoff value to round model prediction probability.

Returns:

List of coordinates [r, c].