I'm a freelance computer-vision developer with a PhD in human-movement science, and I build applied technology for health, physical activity and sports performance. This is the story of why I ended up writing my own multi-camera calibration tool — and why I open-sourced it.
The dependency I couldn't avoid
A lot of my client work rests on the same unglamorous first step: calibrating a rig of USB cameras. Before you can reconstruct a movement in 3D, track a joint or line up several viewpoints, every camera needs its intrinsics (focal length, lens distortion) and its extrinsics (its 6-DoF position and orientation in one shared coordinate frame). Get calibration wrong and everything downstream is wrong — so it kept coming back, project after project, as the quiet dependency I couldn't skip.
Where the existing tools rubbed
The best open-source reference I found was , no pre-recording.
Headless, any device. The service runs in Docker on the machine the cameras are plugged into — no desktop on that host — and you drive everything from a browser on any device on the local network. (Here's the . The dangerous axis-remap math is done for you.
If you already use Caliscope, the TOML output keeps its semantics, so your existing pipelines keep working — I wrote a full
Code & issues: , created by Mac Prible. I use Claude Code (Opus 4.8) to assist me in writing the code.
SOCIAL SHARE CARD GENERATOR