Caliscope
is a GUI-based multicamera calibration package. When the intrinsic (focal length, optical center, and distortion) as well as extrinsic (relative rotation and translation) properties of a set of cameras are known, synchronized frames from those cameras can be used to triangulate landmarks identified across their multiple points of view. With more cameras, this 3D tracking becomes more robust to occlusion and the inevitable errors in 2D landmark tracking and camera property estimates.
While OpenCV provides straightforward functions for the estimation of single camera intrinsic properties as well as estimates of the extrinsic properties of two cameras, there is no straightforward way to estimate extrinsic properties for more than two cameras. Performing this requires bundle adjustment, which demands an extensive series of computational steps and intermediate data tracking.
Caliscope automates this more complex calibration function along with providing visual feedback regarding parameter estimates at each stage of processing. Additionally, there are sample implementations of a Tracker class using Google's Mediapipe that demonstrate the capacity to integrate the full calibration results with landmark tracking tools to achieve 3D pose estimation. While Mediapipe pose estimation has limitations regarding accuracy and precision, it demonstrates a data processing pipeline that can easily integrate more powerful tracking tools as they emerge.
This README provides a general overview and quick guide to install Calicope
on your system. For a more detailed description of the calibration process and workflow, please see our docs.
For a quick sense of how the calibration process works, check out the video walk through of the sample project.
https://github.com/mprib/caliscope/assets/31831778/803a4ce8-4012-4da1-87b9-66c5e6b31c59
Caliscope
was used to calibrate the cameras used in the above recordings (both intrinsic and extrinsic parameters), run 2D landmark estimation across all the videos using Holistic Mediapipe pose estimation, then triangulate the 3D landmark positions based on the camera system calibration.
The animation was created in Blender based on these 3D landmark estimates using a companion project called Rigmarole.
The concurrent video was recorded with a 7 year old t440p laptop, 4 webcams (~$25 each) and another project called MultiWebCam.
Please note that given the size of some core dependencies (OpenCV, Mediapipe, and PySide6 are among them) installation and initial launch can take a while.
# Open Command Prompt and navigate to directory that will hold venv
# this does not need to be the same as where your project workspace is held
cd path\to\your\project
# Create a virtual environment named 'env' using Python 3.10
"C:\Path\To\Python3.10\python.exe" -m venv .venv
# Activate the virtual environment
.\.venv\Scripts\activate
# Your virtual environment is now active.
# You can install using pip
pip install caliscope
# Launch from the command line
caliscope
# Open Command Prompt and navigate to directory that will hold venv
# this does not need to be the same as where your project workspace is held
cd path/to/your/project
# Create a virtual environment named 'venv' using Python 3.10
/path/to/python3.10 -m venv .venv
# Activate the virtual environment
source .venv/bin/activate
# Your virtual environment is now active.
# You can install using pip
pip3 install caliscope
# Launch from the command line
caliscope
Thank you to @calcoloergosum for raising an issue regarding a bug in numpy on MacOS that was causing a crash (https://github.com/mprib/caliscope/issues/622).
It is advised to set the following environment variables on MacOS. If anyone has any additional details related to this or notices that the bug has been fixed, please let me know and I'll take this down.
export MKL_NUM_THREADS=1
export NUMEXPR_NUM_THREADS=1
export OMP_NUM_THREADS=1
png
files for ChArUco calibration boards Automated calculation of camera intrinsic properties from input video
Visualization of distortion model to ensure reasonableness
.trc
file format for use in biomechanical modelling.csv
format with well-labelled headers for straightforward integration with other workflowsThe workflow currently requires you to provide your own synchronized frames or to provide a file that specifies the time at which each frame was read so that caliscope can perform the synchronization itself. There are plans to manage this synchronization automatically through audio files, though that has not yet been implemented.
Google's Mediapipe provides a relatively easy and efficient method for human subject tracking, though for many uses it is limiting. Caliscope has a general Tracker base class that is implemented in a few versions (Pose/Hands/Holistic). This has provided a proof of concept implementation of markerless tracking, though for more robust use the roadmap calls for integration with more powerful tools such as MMPose and DeepLabCut.
To report a bug or request a feature, please open an issue. Please keep in mind that this is an open-source project supported by volunteer effort, so your patience is appreciated.
Post any questions in the Discussions section of the repo.
This project was inspired by FreeMoCap (FMC), which is spearheaded by Jon Matthis, PhD of the HuMoN Research Lab. The FMC calibration and triangulation system is built upon Anipose, created by Lili Karushchek, PhD. Caliscope was originally envisioned as an alternative calibration tool to Anipose that would allow more granular estimation of intrinsics as well as visual feedback during the calibration process. Several lines of of the original Anipose triangulation code are used in this code base, though otherwise it was written from the ground up. I'm grateful to Dr. Matthis for his time developing FreeMoCap, discussing it with me, pointing out important code considerations, and providing a great deal of information regarding open-source project management.
Caliscope is licensed under the permissive BSD 2-Clause license. The triangulation function was adapted from the Anipose code base which is also licensed under the BSD-2 Clause. A primary dependency of this project is PySide6 which provides the GUI front end. PySide6 is licensed under the LGPLv3. Caliscope does not modify the underlying source code of PySide6 which is available via PyPI.