habitat_sim.sensors.noise_models module

Library of Sensor noise models

A library of noise models to close the gap between simulated observations and observations from real sensors.

A noise model can be applied to a sensor by specifying the name of the noise model in the sensor.SensorSpec.noise_model feild. Arguments can be passed to the noise model constructor as keyword arguments using the sensor.SensorSpec.noise_model_kwargs field. For instance, to use the RedwoodDepthNoiseModel with a noise_multiplier of 5

sensor_spec.noise_model = "RedwoodDepthNoiseModel"
sensor_spec.noise_model_kwargs = dict(noise_multiplier=5)

These noise models are commonly the result of contributions from various research projects. If you use a noise model in your research, please cite the relevant work specified by the docummentation

Depth Noise Models

  • Redwood Noise Model for PrimSense depth cameras: RedwoodDepthNoiseModel

Classes

class SensorNoiseModel
Base class for all sensor noise models
class RedwoodDepthNoiseModel
Redwood Noise Model for PrimSense depth cameras
class NoSensorNoiseModel
No noise noise model. Simply returns a copy of the input
class GaussianNoiseModel
class SaltAndPepperNoiseModel
class PoissonNoiseModel
class SpeckleNoiseModel

Functions

def make_sensor_noise_model(name: str, kwargs: typing.Dict[str, typing.Any]) -> SensorNoiseModel
Constructs a noise model using the given name and keyword arguments

Function documentation

def habitat_sim.sensors.noise_models.make_sensor_noise_model(name: str, kwargs: typing.Dict[str, typing.Any]) -> SensorNoiseModel

Constructs a noise model using the given name and keyword arguments

Parameters
name The name of the noise model in the habitat_sim.registry
kwargs The keyword arguments to be passed to the constructor of the noise model