koide3/direct_visual_lidar_calibration

Health v2 · maintenance only

Flags: no_releases no_license

Adoption not part of the score

1507 stars · 220 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

A C++ toolbox for target-less, single-shot extrinsic calibration between LiDAR sensors and cameras, supporting spinning and non-repetitive LiDARs and pinhole, fisheye, and omnidirectional camera models. It uses pixel-level direct LiDAR-camera registration with automatic estimation requiring no initial guess, and integrates with ROS1/ROS2.

Use cases

• calibrate a camera to a LiDAR without a calibration target
• find the extrinsic transform between a point cloud and camera image
• calibrate fisheye or omnidirectional cameras with non-repetitive scan LiDARs
• perform single-shot LiDAR-camera calibration from one data pair
• improve calibration accuracy using multiple LiDAR-camera data pairs
• calibrate robot or vehicle sensors in a ROS1 or ROS2 pipeline

When to choose

• you need LiDAR-camera extrinsic calibration without placing a calibration target
• your setup uses non-standard LiDAR scan patterns or fisheye/omnidirectional cameras
• you have no initial guess for the sensor transform and want an automatic process
• you work in a ROS1/ROS2 environment and want a research-backed (ICRA2023) tool

When to avoid

• you need online, continuous calibration during operation rather than offline single-shot calibration
• you require a license other than MIT or need formal support
• you need LiDAR-inertial or multi-LiDAR calibration instead of LiDAR-camera
• you cannot build C++ dependencies like PCL, GTSAM, and Ceres

Facets

library · maturity stable

computer-vision image-processing math developer-tools robotics autonomous-vehicles computer-vision cpp lidar-camera-calibration target-less-calibration extrinsic-calibration point-cloud-registration ros1 ros2 nid single-shot-calibration algorithms linux ros docker

Member repositories

For agents

markdown · JSON · MCP: product_card(name="koide3/direct_visual_lidar_calibration")