A C++/TensorRT inference module for RangeNet++, enabling fast LiDAR semantic segmentation for robotics applications.
Rangenet Library is an inference module for the RangeNet++ deep learning model, which performs semantic segmentation on LiDAR point cloud data. It provides a C++ interface optimized with TensorRT to classify objects in 3D scans in real-time, enabling applications like autonomous driving and robotic perception.
Robotics researchers and engineers working on autonomous vehicles, semantic SLAM, or LiDAR perception systems who need efficient, real-time semantic segmentation of point clouds.
Developers choose this library for its high-performance C++ implementation, TensorRT acceleration for low-latency inference, and seamless integration with robotics frameworks, making it ideal for real-time applications where speed is critical.
Inference module for RangeNet++ (milioto2019iros, chen2019iros)
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Leverages NVIDIA TensorRT to optimize inference, enabling real-time processing of LiDAR scans on GPU hardware, as highlighted for robotics applications.
Offers a high-performance C++ API tailored for robotics and embedded systems, ensuring low latency crucial for autonomous navigation.
Designed to seamlessly integrate with semantic SLAM pipelines like SuMa++, facilitating advanced robotic perception without extra customization.
Works with pre-trained RangeNet++ models for out-of-the-box semantic segmentation, reducing the need for retraining from scratch.
Only supports TensorRT version 5, with no built-in compatibility for newer versions; adapting requires manual code changes, as admitted in the README referencing issue #9.
Requires precise installation of CUDA, TensorRT 5, and multiple system dependencies, plus the catkin build system, making initial configuration time-consuming and error-prone.
The first inference run takes several minutes to generate a .trt model, delaying immediate testing and deployment in time-sensitive scenarios.