Caliscope calibrates multicamera systems for 3D motion capture. It estimates the intrinsic and extrinsic parameters of every camera, so downstream tools can triangulate 3D positions accurately. Bundle adjustment across 3+ cameras needs a good starting estimate to converge reliably. Caliscope produces that estimate and solves for a quality calibration.
5_cam_demo_small.mp4
We recommend uv for installation. Full instructions (including uv setup and virtual environments) are in the docs.
# Calibration library (intrinsic/extrinsic calibration + triangulation)
uv pip install caliscope
# Desktop app with 3D visualization and pose tracking
uv pip install caliscope[gui]The base install includes the full calibration pipeline as importable Python functions.
Add [gui] for the interactive desktop application, which also includes ONNX pose tracking.
# Launch the app (with the virtual environment activated)
caliscopegit clone https://github.com/mprib/caliscope.git
cd caliscope
uv sync --group dev --extra gui| Feature | What it does |
|---|---|
| Pairwise PnP initialization | Estimates camera positions from stereo pairs chained transitively, so bundle adjustment starts from a reliable point |
| Flexible calibration targets | ChArUco, ArUco, and chessboard targets. A single ArUco marker on a sheet of paper can calibrate a wide volume |
| Mirror board support | A charuco board printed on both sides of a rigid surface links cameras that never share a common view |
| Visual feedback | Inspect distortion models, 3D camera positions, reprojection errors, and world scale accuracy at each stage |
| Outlier filtering | Filter calibration points by reprojection error and re-solve |
| Aniposelib export | Generates camera_array_aniposelib.toml for use with aniposelib-compatible tools |
The base install exposes intrinsic and extrinsic calibration as Python functions with Rich progress bars. See the Scripting API docs for a walkthrough.
Caliscope includes a basic reconstruction pipeline for verifying calibration quality. You can load ONNX pose estimation models (RTMPose, SLEAP, DeepLabCut, or custom) and export 3D trajectories as CSV or TRC (OpenSim).
For full reconstruction workflows, anipose and Pose2Sim are better suited. Caliscope exports an aniposelib-compatible calibration file, making it simple to calibrate here and hand off to those tools.
For a walkthrough with sample data, see the sample project.
To report a bug or request a feature, open an issue. For questions, post in Discussions.
Caliscope was inspired by anipose, created by Lili Karashchuk, PhD, which demonstrated the value of accessible multicamera calibration for the research community.
Caliscope is licensed under the BSD 2-Clause license.