calicam

Camera calibration tool for USB cameras.

Features

• Live streaming from any USB/built-in camera (by index) or pre-recorded video file
• Auto-capture mode (default) — frames are accepted automatically when the board is detected and the camera has moved enough relative to the previous capture
• Running reprojection error — updates after every new frame so you can watch it stabilise in real time
• Dear PyGui GUI — GPU-accelerated, zero-dependency desktop UI
• YAML output matching OpenCV's cv::FileStorage format (compatible with Bonsai, FreeMoCap, etc.)
• JSON output for everything else
• CLI helpers: convert, inspect

Installation

uv tool install calicam

Requires Python 3.10+.

Quick start

calicam gui

# Convert between formats
calicam convert calibration.yaml            # -> calibration.json
calicam convert calibration.json -o cam0.yaml

# Print a summary
calicam inspect calibration.yaml

GUI overview

Reprojection error colour coding: green < 0.5 px · yellow 0.5–1.0 px · red > 1.0 px

YAML format

%YAML:1.0
image_width: 640
image_height: 480
camera_matrix: !!opencv-matrix
   rows: 3
   cols: 3
   dt: d
   data: [ 1.08e+03, 0., 1.93e+02, 0., 1.13e+03, 2.53e+02, 0., 0., 1. ]
distortion_coefficients: !!opencv-matrix
   rows: 5
   cols: 1
   dt: d
   data: [ -1.25, -4.12, -0.005, 0.163, 18.98 ]
reprojection_error: 8.39e-01

Python API

from calicam import Calibrator, CameraSource, save_yaml, load_yaml

# Headless calibration from a video file
with CameraSource("/path/to/video.mp4") as cam:
    calibrator = Calibrator(board_cols=9, board_rows=6, square_size_mm=25.0)
    for frame in cam.frames():
        found, corners, _ = calibrator.detect(frame)
        if found and calibrator.should_auto_capture(corners):
            error = calibrator.add_frame(corners, (cam.width, cam.height))
            if error is not None:
                print(f"  frames={calibrator.n_frames}  RMS={error:.4f}")
        if calibrator.n_frames >= 40:
            break

result = calibrator.calibrate((cam.width, cam.height))
save_yaml(result, "calibration.yaml")

Generating calibration patterns

Print your own board at the correct physical size using the generate command. The --dpi flag controls output resolution (300 is print-ready); the --output / -o flag sets the filename — the extension determines the image format (.png, .tiff, .pdf, …).

# Checkerboard — 9x6 inner corners, 25 mm squares
calicam generate checkerboard --cols 9 --rows 6 --square-size 25 -o board.png

# ChArUco — 5x7 squares, embedded 5x5 ArUco markers
calicam generate charuco --cols 5 --rows 7 --square-size 30 --marker-size 22 -o charuco.png

# Asymmetric circle grid — 4 columns, 11 rows
calicam generate circles --cols 4 --rows 11 --spacing 20 --radius 5 -o circles.png

All commands report the output dimensions in both pixels and millimetres so you can verify the print size before sending to a printer.

Tips

• Use a flat, rigid checkerboard. Print on matte paper, mount on a hard surface.
• Aim for 20–40 frames covering the full image area — tilt and rotate the board.
• A reprojection error below 0.5 px is excellent; below 1.0 px is acceptable.
• If the error stays high, try removing outlier frames and recalibrating (Reset → recapture).