Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/ethz-asl/kalibr/llms.txt

Use this file to discover all available pages before exploring further.

kalibr_camera_validator is a live validation tool that subscribes to ROS camera topics and displays reprojection error statistics in real time. It loads a camera chain calibration from a camchain.yaml file and a calibration target configuration, then shows a per-camera mono view with reprojection error overlays and a stereo-rectified pair view for each camera pair defined in the chain. This tool is useful for verifying a calibration in the field before proceeding to IMU calibration, and for qualitatively assessing how well the camera model fits the optics in live conditions.

​
Synopsis

kalibr_camera_validator --cam <camchain.yaml> --target <target.yaml>

​
Example

kalibr_camera_validator \
  --cam results-camchain.yaml \
  --target aprilgrid.yaml
Point the camera at the calibration target while the validator is running. Reprojection errors are displayed as overlaid statistics in each camera window.

​
Flags

​
--target
string
required
Path to the calibration target configuration YAML file. See target YAML format.
​
--cam
string
required
Path to the camera chain calibration YAML file, typically the camchain.yaml output from kalibr_calibrate_cameras. See camchain YAML format.
​
--verbose
boolean
Enable detailed debug-level logging.

​
Live display windows

The validator opens one OpenCV window per camera and one rectified pair window per overlapping camera pair:
WindowContent
Camera: <topic>Live image with detected corners, reprojected corners (yellow circles), and per-frame reprojection error statistics
Rectified view (cam0 and cam1)Side-by-side stereo-rectified images with horizontal epilines and cross-camera reprojection error statistics

​
Reported statistics

For each camera and each camera pair, the validator displays:
StatisticDescription
mean_x, std_x, max_y, min_xX-component reprojection error statistics (px)
mean_y, std_y, max_y, min_yY-component reprojection error statistics (px)
mean_L2, std_L2, max_L2, min_L2Euclidean (L2) reprojection error statistics (px)
Processing is throttled internally to approximately 2 Hz to keep the display responsive. The validator uses approximate time synchronization (±20 ms) to align frames across cameras.

​
Supported image message types

The validator subscribes to the topic specified in each camera’s camchain.yaml. Both sensor_msgs/Image and sensor_msgs/CompressedImage are supported. If the topic name contains the string compressed, the tool subscribes to CompressedImage automatically.

​
Stereo rectification

For each pair of cameras that share overlapping fields of view (as defined by the chain topology), the validator computes a stereo rectification using the Fusiello-Trucco-Verri algorithm. The rectified images are shown side-by-side. Horizontal epilines confirm alignment: corresponding points should lie on the same horizontal line in both rectified images if the calibration is accurate.

Build docs developers (and LLMs) love

Get started for freeTalk to us