How to install depth_clustering on Ubuntu 18.04?

Install dependencies like libopencv-dev, libqglviewer-dev-qt5, freeglut3-dev, and qtbase5-dev via apt, then build using catkin or CMake as per the README. Note that you might need to handle GoogleTest binaries separately.

Depth clustering vs Euclidean clustering for LiDAR data?

Depth clustering is specialized for real-time segmentation of sparse Velodyne scans using sensor geometry for efficiency, while Euclidean clustering is a general-purpose method that may be slower for online applications. Depth clustering excels in speed but is less flexible with non-Velodyne data.

Can depth_clustering work with non-Velodyne LiDAR data?

It's primarily designed for Velodyne sensors; using other LiDAR data requires converting to supported formats (like KITTI .bin or Moosmann .png) and may involve tweaking the algorithm, as it relies on Velodyne-specific beam configurations.

How to visualize segmentation results in depth_clustering?

Run the qt_gui_app binary from the build folder, click the OpenFolder button, and load data files (e.g., .png for Moosmann's dataset). The GUI provides interactive controls to navigate and view segmented objects outlined in orange.

What datasets are supported by depth_clustering?

It supports the KITTI dataset, Moosmann's Velodyne SLAM dataset, and ROS topics. Examples in the repository show how to run on these, with binaries like show_objects_moosmann for specific data formats.

Is depth_clustering suitable for real-time autonomous driving?

Yes, it's optimized for real-time online operation and has been validated in academic papers for autonomous systems. However, performance depends on hardware and data rates, so testing with your specific setup is recommended.

Open-Awesome

depth_clustering

MITC++v2.0.0

Fast and robust algorithm for segmenting Velodyne LiDAR point clouds into objects for autonomous driving applications.

GitHub

1.3k stars384 forks0 contributors

What is depth_clustering?

Depth Clustering is a C++ library for fast segmentation of 3D point clouds generated by Velodyne LiDAR sensors. It processes sparse laser scan data to identify and separate objects in real-time, which is essential for autonomous vehicles and robotics applications. The algorithm is specifically optimized for online operation with Velodyne's rotating multi-beam sensors.

Target Audience

Autonomous vehicle developers, robotics researchers, and computer vision engineers working with LiDAR point cloud data who need real-time object segmentation capabilities.

Value Proposition

Developers choose Depth Clustering for its proven performance with Velodyne sensors, real-time processing capabilities, and robust segmentation results that have been validated through academic publications and practical applications in autonomous systems.

Overview

:taxi: Fast and robust clustering of point clouds generated with a Velodyne sensor.

Use Cases

Best For

Real-time object detection from Velodyne LiDAR data in autonomous vehicles
Segmenting sparse 3D point clouds for robotics perception systems
Processing KITTI dataset or similar autonomous driving datasets
ROS-based robotic systems that need online LiDAR segmentation
Research projects involving 3D point cloud analysis and segmentation
Developing perception pipelines for self-driving cars using Velodyne sensors

Not Ideal For

Dense point cloud segmentation from structured-light or time-of-flight sensors
Offline batch processing where maximum accuracy is prioritized over speed
Projects using non-Velodyne LiDAR sensors with different data formats
Teams without expertise in C++, catkin, or ROS ecosystems

Pros & Cons

Pros

Real-Time Performance

Optimized for fast online operation, enabling real-time processing of streaming LiDAR data as required for autonomous vehicles, with examples showing live segmentation on datasets.

Velodyne Sensor Compatibility

Works with all Velodyne models (16, 32, 64 beam) as stated in the README, ensuring broad applicability in standard LiDAR setups for robotics and autonomous driving.

ROS Integration

Provides ROS nodes for seamless integration into robotic systems, allowing easy deployment in existing perception pipelines with support for ROS topics.

Visualization Tools

Includes a Qt-based GUI for interactive visualization of segmentation results, aiding in debugging and analysis with features like folder loading and navigation controls.

Cons

Complex Dependency Setup

Requires installation of multiple libraries (OpenCV, QGLViewer, FreeGLUT, QT) with version-specific commands for different Ubuntu systems, making initial deployment cumbersome.

Sensor Specificity

Primarily optimized for Velodyne sensors; adapting to other LiDAR brands may require significant modification of data input formats and algorithm parameters.

Sparse Data Focus

Designed for sparse Velodyne scans, so it may not perform optimally with dense point clouds from other sources, limiting versatility in varied applications.

Frequently Asked Questions

Related Projects

GitHub repository

Open3D: A Modern Library for 3D Data Processing

Draco 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.

Stars7,247

Forks1,053

Last commit20 days ago

mmdetection3d

OpenMMLab's next-generation platform for general 3D object detection.

Stars6,399

Forks1,758

Last commit1 year ago

PCDet

OpenPCDet Toolbox for LiDAR-based 3D Object Detection.

Stars5,548

Forks1,446

Last commit6 months ago

Community-curated · Updated weekly · 100% open source

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.

Submit a project Star on GitHub