mirage.sim package

Submodules

mirage.sim.interp module

class mirage.sim.interp.MLPBrightnessModel(obj: SpaceObject | None = None, brdf: Brdf | None = None, path: str | None = None, use_engine: bool = True, layers: tuple[int] = (150, 50, 150), **engine_kwargs)

Bases: object

Predicts brightness by evaluating full non-convex rendered brightness over a 6D cloud of lighting and illumination unit vectors, used to train a Multi-Layer Perceptron (MLP)

brightness(svb: ndarray, ovb: ndarray, output_as: str | None = None) → ndarray

True brightness function, evaluates using the LightCurveEngine renderer or a convex BRDF method

Parameters:

• svb (np.ndarray [n,3]) – Sun unit vector in the body frame

• ovb (np.ndarray [n,3]) – Observer unit vectors in the body frame

• output_as (str) – Values to return, ‘irradiance’ or ‘magnitude’, defaults to None (uses self.train)

Returns:

Unit distance irradiance in given observation geometry [W/m^2]

Return type:

np.ndarray

eval(svb: ndarray, ovb: ndarray, eval_mode_pref: str = 'sklearn') → ndarray

Evaluates the brightness model with the given observer and sun unit vectors in the body frame

Parameters:

• svb (np.ndarray [n,3]) – Sun vector in body frame, length irrelevant

• ovb (np.ndarray [n,3]) – Observer vector in body frame, length irrelevant

• eval_mode_pref (str) – Evaluation mode preference, defaults to ‘sklearn’

Raises:

AttributeError – If the model has not been built yet

Returns:

Predicted normalized irradiances [W/m^2]

Return type:

np.ndarray [n,]

gen_train_dataset(num: int) → Tuple[ndarray, ndarray, ndarray, ndarray]

load_from_file(path: str) → None

Loads the model from a saved pickle file

Parameters:

path (str) – Path to the pickle file containing the saved model

save_to_file(directory: str = '', save_as_format: str = 'onnx') → None

Saves the model to the directory provided

Parameters:

• directory (str, optional) – Directory to save model in, defaults to “”

• save_as_format (str) – Format to save the file to, either ‘onnx’ or ‘sklearn’, defaults to ‘onnx’

train(num_train: int) → None

Builds the MLP model, enables self.eval(points) after completion

Parameters:

num_train (int, optional) – Number of data points to train on

Module contents