finds.dev

hku-mars/r3live

A Robust, Real-time, RGB-colored, LiDAR-Inertial-Visual tightly-coupled state Estimation and mapping package

View on GitHub

★ 2,481C++GPL-2.0updated Dec 2025

// the take

R3LIVE is a ROS1 package from HKU-MARS that fuses LiDAR, IMU, and camera data to produce real-time, RGB-colored 3D point cloud maps. It's aimed at robotics/SLAM researchers with Livox hardware who want colored maps rather than just geometric ones — not a general-purpose SLAM library for arbitrary sensor setups.

Tightly-coupled LIO+VIO fusion built on top of already-proven components (FAST-LIO for odometry, ikd-Tree for kNN search) rather than a from-scratch reimplementation, and the paper backs up real-time performance in LiDAR-degenerate and texture-poor scenes with actual benchmark data. It also ships the full offline pipeline for mesh reconstruction and texturing, so you get usable 3D assets out the other end instead of just a raw colored point cloud.

GPLv2 with an explicit carve-out that commercial use requires contacting the authors for a separate license — a real blocker if you're building a product on top of it. It's hard-wired to Livox LiDARs via a custom-patched livox_ros_driver fork; using a spinning LiDAR means manually editing LiDAR_front_end.cpp yourself, and the authors only give one Ouster example with VIO disabled. OpenCV version mismatches are a recurring crash source per their own issue tracker, and running it at all requires careful manual extrinsic calibration between LiDAR, camera, and IMU — there's no auto-calibration path. The authors themselves describe it as not fully stable, and it's ROS1/catkin, so anyone on ROS2 is stuck either porting it or running a compatibility layer.

We read the README, the directory tree and a key source file in October 2026. Nobody pays to be listed. How we judge →

// at a glance

Stars
2,481
Forks
486
Language
C++
License
GPL-2.0
Last commit
4 Dec 2025

// nearby

If r3live isn't quite it.

All repos →