SmartFactorParams - GTSAM Docs

The SmartProjectionParams class configures smart projection factors. It controls their linearization, triangulation, and degeneracy behavior.

import gtsam
import numpy as np

from gtsam.symbol_shorthand import X

Configure SmartProjectionParams¶

We construct SmartProjectionParams with Jacobian-Q linearization and zero-information handling for degenerate triangulation. The setter methods then tune rank, epipolar-initialization, landmark-distance, and dynamic-outlier thresholds for the intended scene.

params = gtsam.SmartProjectionParams(
    gtsam.LinearizationMode.JACOBIAN_Q,
    gtsam.DegeneracyMode.ZERO_ON_DEGENERACY,
    False,
    False,
    1e-5,
)
params.setRankTolerance(1e-8)
params.setEnableEPI(False)
params.setLandmarkDistanceThreshold(1e4)
params.setDynamicOutlierRejectionThreshold(3.0)
params.print("configured smart-factor parameters")

linearizationMode: 2
   degeneracyMode: 1
rankTolerance = 1e-08

enableEPI = 0

landmarkDistanceThreshold = 10000

dynamicOutlierRejectionThreshold = 3

useLOST = 0

noise model

Apply the parameters to a smart factor¶

Pass the configured object to a smart projection factor constructor. After adding two consistent camera observations, linearizeToJacobian() returns a JacobianFactorQ, confirming that the selected linearization mode took effect.

calibration = gtsam.Cal3_S2(500.0, 500.0, 0.0, 320.0, 240.0)
cameras = [
    gtsam.PinholeCameraCal3_S2(gtsam.Pose3(), calibration),
    gtsam.PinholeCameraCal3_S2(
        gtsam.Pose3(gtsam.Rot3(), np.array([1.0, 0.0, 0.0])),
        calibration,
    ),
]
landmark = np.array([0.0, 0.0, 5.0])
factor = gtsam.SmartProjectionFactorPinholeCameraCal3_S2(
    gtsam.noiseModel.Isotropic.Sigma(2, 1.0), params
)
values = gtsam.Values()
for index, camera in enumerate(cameras):
    factor.add(camera.project(landmark), X(index))
    values.insert(X(index), camera)

linear_factor = factor.linearizeToJacobian(values)
assert isinstance(linear_factor, gtsam.JacobianFactorQ112)
assert np.isclose(factor.totalReprojectionError(factor.cameras(values)), 0.0)
print("linear factor:", type(linear_factor).__name__)

linear factor: JacobianFactorQ112

When to use it¶

Use SmartProjectionParams whenever a smart projection factor needs non-default triangulation, degeneracy, outlier, or linearization behavior. Start with the default Hessian mode unless the downstream solver specifically consumes Jacobian factors, and choose thresholds in the same scale as the scene and measurement noise.

SmartProjectionFactor shows the factor itself, while JacobianFactorQ explains the linear factor returned by this example.

Source¶

SmartFactorParams.h

AI assistance caveat¶

AI was used to help draft this documentation, and inaccuracies could be present.