xuankuzcr

xuankuzcr / FAST-Calib2

Public

A Robust LiDAR-camera Calibration Tool for Large-spot LiDARs.

48
3
89% credibility
Found May 25, 2026 at 65 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
C++
AI Summary

FAST-Calib2 is a research tool that helps robots and vehicles calibrate their LiDAR and camera sensors so they can understand the world together. It uses a special calibration board with reflective ring patterns and visual markers. The software automatically detects this board in sensor data, extracts key reference points, and calculates the precise spatial relationship between the two sensors. It works with many types of LiDAR sensors (both solid-state and mechanical) and can combine data from multiple positions for highly accurate calibration results.

How It Works

1
🤖 You have a robot with both eyes

Your robot or vehicle is equipped with a LiDAR sensor (like a spinning radar) and a camera that need to work together as a unified perception system.

2
🎯 You print the special calibration board

You create the custom calibration target with four reflective ring patterns and four visual markers—this is the key to making your sensors see the same thing.

3
📸 You capture data from different angles

You place the calibration board in front of your robot and record sensor data from at least three different positions, moving the board each time.

4
✨ The software automatically finds the board

Without any manual tweaking, the software automatically locates the calibration board in your LiDAR point cloud and detects the visual markers in your camera images.

5
🔍 It extracts the center points

For each position, the software finds the exact center of each ring pattern in the LiDAR data and each marker in the camera data.

6
Choose your calibration approach
⚡
Single scene

Fast calibration using just one position—great for checking if everything is working

🎯
Multiple scenes

Combine three or more positions for more robust and accurate results

🎉 You get the precise relationship between your sensors

The software outputs the exact rotation and translation that tells your robot how its LiDAR and camera are positioned relative to each other, ready for accurate 3D perception.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 65 to 48 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is FAST-Calib2?

FAST-Calib2 is a LiDAR-camera extrinsic calibration tool written in C++ that estimates the rigid transformation between a LiDAR and camera sensor. It uses a custom-designed reflective annular calibration target with four concentric rings, which the LiDAR detects via reflectivity and the camera detects via ArUco markers. The tool handles both solid-state LiDARs (Livox Mid360, Avia) and mechanical LiDARs (Hesai JT128, Ouster) with separate detection pipelines optimized for each type's point cloud characteristics.

Why is it gaining traction?

The main draw is automatic board detection -- no manual ROI tuning required. The system auto-locates the calibration target using high-reflectivity clustering and geometric consistency checks against the known target layout. For solid-state LiDARs with noisy point clouds, the dual-path approach (single circle fitting with fallback to boundary-based concentric circle fitting) provides robustness. The tool also outputs quality metrics: center-to-center geometry error and annulus radius consistency checks, so you know when calibration succeeded versus when it produced garbage.

Who should use this?

Robotics engineers integrating LiDAR-camera sensor suites who need extrinsic calibration without manual parameter tweaking. Perception researchers working with solid-state LiDARs like Livox devices, where traditional circular-hole targets struggle with center extraction errors. Teams running automated calibration pipelines that benefit from built-in quality validation rather than post-hoc inspection.

Verdict

This is a specialized tool for a specific calibration workflow -- if you need LiDAR-camera extrinsic calibration with reflective annular targets, this handles the full pipeline cleanly. The 48 stars and academic origin suggest a niche but active project. The credibility score of 0.8999999761581421% indicates reasonable provenance but limited community validation. Documentation covers the workflow adequately, though expect to read the code for edge cases. Consider it production-ready for established calibration targets, experimental for custom configurations.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.