January 28, 2024: The package has now been renamed caliscope
(previously pyxy3d
). I've recently completed some major refactors that enable a pre-recorded workflow, split out the sychronized webcam recording into its own project (multiwebcam), and have started the ball rolling on a submission for the Journal of Open Source Software. I'd been wanting to change the name for awhile, and with all this transition it seems like the right time. In the coming days I will do a final update to PyPI for pyxy3d which will cause it to just pass through the caliscope package as this seems like the most robust way to part ways with that legacy package.
Caliscope
is a GUI-based multicamera calibration package that integrates with 2D landmark tracking tools to produce low-cost, open-source motion capture.
Please see our docs for details about installation, project setup, and general workflow.
ballet_demo.mp4
The above was created using Caliscope, a 7 year old t440p laptop, and 4 webcams (~$25 each). This includes camera calibration, recording of synchronized frames (720p @ 24 fps), landmark detection, and point triangulation. Note that the webcam recording functionality is not in the current version, though will be restored in the future. Animated rig creation was done using an early stage Blender add-on project called Rigmarole
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
.\env\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
- Once the GUI launches, navigate to File->New/Open Project and create a folder to hold your project
- A basic project structure will be created here
- Define a Charuco calibration board via the Charuco tab and print it out, fixing it to something flat
- Record footage for the calibration according to the guidelines for the intrinsic and extrinsic calibrations.
- Record synchronized motion capture trials
- A companion project (multiwebcam) has been set up to facilitate this though is still in early stages
- Store video files within the project folder and reload the workspace
- Run autocalibration on all cameras within the Cameras tab
- Run "Calibration Capture Volume" from the Workspace Tab
- Set origin within the capture volume (optional but helpful)
- Run post-processing on individual recordings to generate 3D trajectory output
- Easy creation of
png
files for ChArUco calibration boards - board definition can be changed across intrinsic and extrinsic calibration allowing greater flexibiltiy
-
Automated calculation of camera intrinsic properties from input video
- Optical Center
- Focal Length
- Lens Distortion
-
Visualization of distortion model to ensure reasonableness
- Automated bundle adjustment to estimate 6 DoF relative position of cameras
- Visualizer to inspect the estimates from the bundle adjustment
- Setting of the World Origin within the visualizer to simplify data processing
- Tracker API for integrating alternate tracking methods
- 3 sample implementations with Google Mediapipe (Hands/Pose/Holistic)
- Automated application of landmark tracking to synchronized videos
- Triangulation of 3D landmark position based on calibrated cameras
- Gap-filling and butterworth filtering to smooth trajectory estimates
- output to
.trc
file format for use in biomechanical modelling - output to tidy
.csv
format with well-labelled headers for straightforward integration with other workflows - companion project Rigmarole in development to facilitate creation of animated rigs in Blender
The 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.