Cross Validate

SUMMARY

cross_validate() computes the mean reprojection error over n_splits held-out folds of a set of images, the core metric behind run()'s held_out_reprojection_error.

SUPPORTED TARGETS

Works with any target, ChessboardTarget, CharucoTarget, or ArucoTarget, it only runs intrinsic calibration per fold.

UNITS

Input images are BGR ndarrays. Returns the mean held-out reprojection error as a float in pixels.

The Skill

error = bench.cross_validate(images, target, options, n_splits=5, seed=0)

Skill	Returns	Description
cross_validate(images, target, options, n_splits=5, seed=0)	float	Mean reprojection error over n_splits held-out folds of images.

Parameter	Type	Description
images	list[ndarray]	BGR calibration images to split into folds.
target	Target	Calibration target to detect.
options	IntrinsicOptions	Intrinsic solve configuration, see IntrinsicOptions.
n_splits	int	Number of cross-validation folds. Default 5.
seed	int	Makes the fold shuffle deterministic across runs. Default 0.

The Code

from telekinesis.axon import CalibrationBenchmark, IntrinsicOptions
from telekinesis.axon.targets import CharucoTarget

target = CharucoTarget(squares_x=6, squares_y=9, square_length=0.012, marker_length=0.009)

bench = CalibrationBenchmark()
error = bench.cross_validate(images, target, IntrinsicOptions(), n_splits=5, seed=0)