Run the checkerboard example
Calibrate the supplied synthetic views, check the report, and plot training and held-out validation errors.
Get the files and start practiceIntermediate · Hands-on
Camera Calibration:
From Pixels to Metric Geometry
Turn image points into a camera model you can test. Build an OpenCV calibration pipeline, inspect its errors, and decide where its results can be used.
Start with two open lessons. Confirm your email to unlock the rest, free. The newsletter is optional.
Edit Python and explore its results in every lesson. Python runs in your browser; no installation or camera is needed.
- Core path
- 13 lessons · 7 hr 25 min
- Optional lessons
- 2 lessons · 1 hr 20 min
- Final project
- A checked report and a reasoned decision about its intended use
Change depth to see perspective in action.
Who this is for
For Python and OpenCV developers working on measurement, pose estimation, or image overlays
You should already know
- Python functions, arrays, and files
- Basic NumPy indexing and matrix notation
- How to load an image with OpenCV
By the end, you'll have
- A runnable checkerboard calibration pipeline
- Plots of projection errors and checks on images excluded from fitting
- A validated JSON report with the settings, results, and limits another engineer needs
Your progress
Your saved course
0 of 13 core lessons
0 of 2 optional lessons
Course curriculum
From camera model to final decision
Follow modules 00-11, then complete the final project. After the final project, explore pose estimation (PnP) and stereo depth in the optional lessons.
Loading curriculum…
Final project
Decide whether a calibration is ready to use
The final project checks the JSON format, numeric values, model settings, image counts, held-out validation results, and stated limits. You then decide whether to approve the calibration and explain why.
Open the final projectContractcamera-calibration-report/v1
AutomatedRequired fields · valid numbers · consistency checks
Your decisionResults · limits · approve or reject
Practice and reference
Run the pipeline and understand its limits
Report format
Review the JSON Schema and sample report before you load a calibration report into another application.
Open JSON SchemaHow the concepts connect
See how the modules build on one another, or browse them in course order.
Open concept mapScope and limitations
The runnable example uses a straightforward checkerboard and Brown-Conrady model. The PnP exercise draws axes on a supplied checkerboard image. Fisheye and stereo are conceptual extensions with model assumptions and reference APIs. Before you use any result, test it with the actual camera, lens, focus, resolution, image processing, and working region.