GitHub repository
A LiDAR-based tool for constructing static maps by removing dynamic points from point cloud sequences.
What is GitHub repository?
Removert is a LiDAR processing tool that constructs static maps by removing dynamic points from sequences of 3D point clouds. It addresses the challenge of creating clean, reliable maps for robotics and autonomous navigation by filtering out moving objects like vehicles and pedestrians. The tool operates offline on pre-recorded data, using a multiresolution range image approach to efficiently separate static and dynamic elements.
Robotics researchers and engineers working on SLAM, autonomous vehicles, or mapping systems who need to generate static maps from LiDAR data. It is particularly useful for those processing datasets like KITTI or MulRan.
Removert offers a specialized, reversible method for dynamic point removal that is more precise than online filters, supports various LiDAR configurations, and integrates with ROS for ease of use. Its offline batch processing ensures high-quality map construction without real-time constraints.
Overview
Remove then revert (IROS 2020)
Use Cases
Best For
- Constructing static maps from urban LiDAR datasets like KITTI
- Removing dynamic objects (e.g., cars, pedestrians) from point cloud sequences
- Enhancing SLAM systems by providing cleaner input maps
- Offline batch processing of large-scale LiDAR data
- Combining with deep learning methods for improved dynamic removal
- Research in robotics mapping and change detection
Not Ideal For
- Real-time autonomous systems requiring immediate dynamic point filtering during operation
- Projects without access to pre-processed LiDAR scans and accurate SLAM pose estimates
- Commercial applications restricted by the non-commercial CC BY-NC-SA 4.0 license
Pros & Cons
Pros
Uses a 'remove-then-revert' approach to carefully preserve static structures while filtering dynamic points, ensuring high accuracy as detailed in the IROS 2020 paper.
Employs range images at multiple resolutions to balance detection accuracy and computational speed, making it effective for large datasets like KITTI and MulRan.
Seamlessly works with ROS for parameter handling and visualization, with provided launch files for easy setup on datasets like KITTI, despite operating offline.
Designed to complement online methods like LiDAR-MOS, enabling hybrid approaches where Removert handles finer offline cleaning after coarse deep learning removal.
Cons
Cannot process data in real-time, restricting use to post-processing scenarios and requiring full datasets to be available upfront, as admitted in the README.
Requires converting raw LiDAR data into specific binary scan formats and pose text files (KITTI-style), adding significant preprocessing overhead before use.
Key components like the revert step are omitted, and full sequence automation is pending, reducing out-of-the-box functionality as noted in the TODO list.
Licensed under CC BY-NC-SA 4.0, which prohibits commercial use without permission, limiting adoption in industry projects and deployments.
Frequently Asked Questions
Related Projects
GitHub repositoryOpen3D: A Modern Library for 3D Data Processing
CartographerCartographer is a system that provides real-time simultaneous localization and mapping (SLAM) in 2D and 3D across multiple platforms and sensor configurations.
DracoDraco is a library for compressing and decompressing 3D geometric meshes and point clouds. It is intended to improve the storage and transmission of 3D graphics.
mmdetection3dOpenMMLab's next-generation platform for general 3D object detection.
Found a gem we're missing?
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.