deepblink.data module¶
List of functions to handle data including converting matrices <-> coordinates.
-
deepblink.data.absolute_coordinate(coord_spot: Tuple[numpy.float32, numpy.float32], coord_cell: Tuple[numpy.float32, numpy.float32], cell_size: int = 4) → Tuple[numpy.float32, numpy.float32][source]¶ Return the absolute image coordinate from a relative cell coordinate.
Parameters: - coord_spot – Relative spot coordinate in format (r, c).
- coord_cell – Top-left coordinate of the cell.
- cell_size – Size of one cell in a grid.
Returns: Absolute coordinate.
-
deepblink.data.get_coordinate_list(matrix: numpy.ndarray, image_size: int = 512, probability: float = 0.5) → numpy.ndarray[source]¶ Convert the prediction matrix into a list of coordinates.
NOTE - plt.scatter uses the x, y system. Therefore any plots must be inverted by assigning x=c, y=r!
Parameters: - matrix – Matrix representation of spot coordinates.
- image_size – Default image size the grid was layed on.
- probability – Cutoff value to round model prediction probability.
Returns: Array of r, c coordinates with the shape (n, 2).
-
deepblink.data.get_prediction_matrix(coords: numpy.ndarray, image_size: int, cell_size: int = 4, size_c: int = None) → numpy.ndarray[source]¶ Return np.ndarray of shape (n, n, 3): p, r, c format for each cell.
Parameters: - coords – List of coordinates in r, c format with shape (n, 2).
- image_size – Size of the image from which List of coordinates are extracted.
- cell_size – Size of one grid cell inside the matrix. A cell_size of 2 means that one cell corresponds to 2 pixels in the original image.
- size_c – If empty, assumes a squared image. Else the length of the r axis.
Returns: The prediction matrix as numpy array of shape (n, n, 3) – p, r, c format for each cell.