tfl.conditional_pwl_calibration.default_keypoint_output_parameters

Helper creating default keypoint_output_parameters.

tfl.conditional_pwl_calibration.default_keypoint_output_parameters(
    num_keypoints: int,
    units: int = 1,
    monotonicity: str = 'none',
    is_cyclic: bool = False,
    clamp_min: bool = False,
    clamp_max: bool = False,
    derived_missing_output: bool = False
) -> Optional[tf.Tensor]

Primarily used for testing.

num_keypoints number of keypoints for inputs.
units number of parallel calibrations on one input.
monotonicity none or increasing, monotonicity of the calibration.
is_cyclic whether the calibration is cyclic. Only works if monotonicity == none.
clamp_min whether the leftmost keypoint should be clamped. Only works if monotonicity == increasing.
clamp_max whether the rightmost keypoint should be clamped. Only works if monotonicity == increasing.
derived_missing_output whether to reserve a placeholder for the missing output value.

A tensor with a shape of (1, units, output_param_size).

ValueError if parsing failed.