v2.0.0 Major rewrite and breaking changes (READ CHANGELOG)
[2.0] - 09.01.2021
2.0 is a rewrite of a lot of the API and requires multiple changes to legacy code using this library.
Please refer to the migration guide for more information.
During this refactoring multiple bugs were fixed as well.
Therefore, it is highly suggested upgrading to the new version, even if it takes some work!
Added
- A new
calibrate_dfmethod for theCalibrationInfothat can calibrate df directly. - It is now possible to define which
CalibrationInfosubclass should be used by theFerrarisCalibration - A set of "management" functions to save, find, and load IMU calibrations of multiple sensors
- The ability to add a custom comment to a
CalibrationInfo - The user is now forced to provide the units of the input data to avoid applying calibrations that were meant for unit
conversion. - Applying a calibration now checks if the units of your data match with the input unit of the calibration.
- The export format of calibration-info objects is now versioned.
This helps to make changes to the format in the future while still supporting old exports.
See the migration guide for more information. - Helper functions to load "legacy" calibration info objects. (
imucal.legacy)
Changed
FerrarisCalibrationhas a new interface.
Instead of providing all calibration and data related parameters in the__init__, the__init__is now only used
to configure the calibration.
The data and all data related parameter are now passed to thecomputemethod (replacescompute_calibration_matrix)- Using
from_...constructors on a subclass ofCalibrationInfodoes not search all subclasses ofCalibrationInfo
anymore, but only the subclasses (and the class itself), it is called on.
For example,FerrarisCalibrationInfo.from_jsonwill only consider subclasses ofFerrarisCalibrationInfo, but not
other subclasses ofCalibrationInfo. - The short hand "gyro" is not replaced with "gyr" in all parameter and variable names.
This might cause an issue when loading old calibration files.
Deprecated
Removed
FerrarisCalibrationdoes not have anyfrom_...constructors anymore.
The functionality of these constructors can now be accessed via theferraris_regions_from_...helper functions.- It is not possible anymore to calibrate the acc and gyro separately.
No one was using this feature, and hence, was removed to simplify the API.
Fixed
Migration Guide
-
The main change is how
FerrarisCalibrationandTurntableCalibrationare used.
Before you would do:from imucal import FerrarisCalibration cal, section_list = FerrarisCalibration.from_interactive_plot(data, sampling_rate=sampling_rate) cal_info = cal.compute_calibration_matrix()
Now you need to first create your Ferraris sections and then provide them as arguments for the
computemethod:from imucal import FerrarisCalibration, ferraris_regions_from_interactive_plot sections, section_list = ferraris_regions_from_interactive_plot(data) cal = FerrarisCalibration() cal_info = cal.compute(sections, sampling_rate_hz=sampling_rate, from_acc_unit="m/s^2", from_gyr_unit="deg/s")
Note, that you are also forced to provide the units of the input data.
We always recommend to first turn your data into the same units you would expect after the calibration and then using
the calibrations as refinement. -
CalibrationInfot.calibratenow requires you to specify the units of your data and validates that they match the
units expected by the calibration.
You need to add the parametersacc_unitandgyr_unitto all calls to calibrate.
Note, that older calibrations will not havefromunits and hence, can not perform this check.
In this case you can set the units toNoneto avoid an error.
However, it is recommended to recreate the calibrations with properfromunits. -
If you were using
from_json_filebefore, double check, if this still works for you, as the way the correct baseclass
is selected have been chosen.
In any case, you should consider to useimucal.management.load_calibration_infoinstead, as it is more flexible. -
If you were using any parameters or package variables, that contained the short hand
gyro, replace it withgyr.
Note that this also effects the exported calibration files.
You will not be able to load them unless you replacegyro_unitwithgyr_unitin all files. -
The
CalibrationInfoobjects now have more fields by default.
To avoid issues with missing values, we highly recommend recreating all calibration files you have using the original
session data and the most current version ofimucal.
Alternatively you can use the functions provided inimucal.legacyto load the old
calibration.
Then you can modify the loaded calibration info object and save it again to replace the old calibration.