/pointpillar-ros-node

Use PointPillar with your ROSbags.

Primary LanguagePythonMIT LicenseMIT

PointPillar-ROS-Node

This repository contains a ROS node that utilizes the PointPillar neural network model (MMDetection3d) to process point cloud data from ROSbags. The repository is specifically designed to work in a ROS 1 environment.

Please note that the instructions provided assume a ROS Noetic environment. The repo is tested with Ubuntu 20.

Prerequisites

Before using this repository, please ensure that you have the following prerequisites installed:

  • CUDA 11.6 and MMDetection3D library:
  • ROS Noetic
  • jsk-recognition-msgs, jsk-rviz-plugins:
    • sudo apt-get install ros-noetic-jsk-recognition-msgs ros-noetic-jsk-rviz-plugins

Additionally, you will need to download the required models weight and config files from model zoo and place them in directory: pointpillar_ros/model/pointpillar/. Update model weight and confi files' names in launchfile.

Usage

To use this code, follow the steps below:

  1. Clone this repository to your catkin_ws/src directory.
  2. Update the point cloud topic and model names in launchfile.
  3. Build the package:
    catkin config --install
    catkin build pointpillar_ros
  4. Launch the ROS node by executing the following commands:
    source install/setup.bash # or source install/pointpillar_ros/setup.bash
    roslaunch pointpillar_ros pointpillar_ros.launch
  5. Launch RViz, visualize the detection bounding boxes.

This is a screenshot of detection result:

Troubleshooting

If you encounter a ModuleNotFoundError related to the netifaces module, you can resolve it by installing the module using pip pip install netifaces.

Hope this helps! If you have any further questions, feel free to ask.