Need help with crisp?
Click the “chat” button below for chat support from the developer who created it, or find similar developers for support.

About the developer

142 Stars 72 Forks GNU General Public License v3.0 66 Commits 5 Opened issues


Camera-to-IMU calibration and synchronization toolbox

Services available


Need anything else?

Contributors list

# 302,601
64 commits

Camera-to-IMU calibration toolbox

This toolbox provides a python library to perform joint calibration of a rolling shutter camera-gyroscope system.

Given gyroscope and video data, this library can find the following parameters

  • True gyroscope rate
  • Time offset
  • Rotation between camera and gyroscope coordinate frames
  • Gyroscope measurement bias

If you use the package for your work, please cite the following paper

Ovrén, H and Forssén, P.-E. "Gyroscope-based video stabilisation with auto-calibration." In 2015 IEEE International Conference on Robotics and Automation (ICRA) (pp. 2090–2097). Seattle, WA

Can I use these methods for my application?

The calibration methods in this package assumes the following

  • Your camera is calibrated, including known readout time
  • The camera frame rate is constant, and known
  • The gyroscope frame rate is constant, and approximately known (within a few Hz, or percent)

If the video and gyroscope data are not uniformly sampled, but you have access to somewhat reliable timestamps, then you can still use the method if you resample the data to be uniform. By "reliable" we mean timestamps without drift, and no (or negligble) jitter.

Changes from 1.0

The 2.0 version of crisp features a new fully automatic calibrator. This means that there is no compelling reason to use the semi-manual methods in the previous version of crisp. Therefore the old example scripts have been removed, and the old functions are not imported into the module namespace. No old functions have been removed, so if you want to use them they are still available in submodules.


To use the package you need the following Python packages:

  • NumPy
  • SciPy
  • OpenCV
  • matplotlib

The easiest way is to install from PyPI:

$ pip install crisp

If you want to build the package from source, you also need the Cython package. To build and install the

module just run the following commands:
$ python build
$ python install

For a user-only installation add

to the install command.


The gyroscope and video data are first loaded into a stream object (

, and a subclass of
respectively). To be able to understand how points are mapped from the real world to the image, the video stream also need a
(-subclass) instance.
import crisp

gyro = crisp.GyroStream.from_data(some_data_array) camera_model = crisp.AtanCameraModel(...) # One specific choice of camera model video = crisp.VideoStream.from_file(camera_model, video_file_path)

We then tie the streams together using a

instance. Since the calibration proces need to have estimates of the time offset and relative rotation, these are first estimated using the
member. This initialization only requires that you give an approximate gyroscope sample rate (in Hz).
calibrator = crisp.AutoCalibrator(video, gyro)
result = calibrator.calibrate() # Dict of calibrated parameters

Initialization and calibration errors can be caught by handling


Example scripts

We bundle one example script
which shows how to use the library with the data in our dataset ( This is the same dataset that was used to produce the above mentioned ICRA 2015 paper.


  • For any questions regarding the method and paper, please send an e-mail to [email protected]
  • For issues about the code, you are welcome to either use the tools (issue reporting, etc.) provided by GitHub, or send an e-mail.


All code in this repository is licensed under the GPL version 3.

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.