pcdet_ros2

This is a ROS 2 Wrapper for [OpenPCDet].

Test System

• Computation: i9-12900K, GPU 4080
• System Setup:
  • Ubuntu 22.04, ROS 2 Humble
  • CUDA 11.7, CuDNN 8.5.0.96
  • Python 3.10, PyTorch 2.0
• OpenPCDet Version: v0.6.0

Status

• Successfully Tested with Kitti 3D Object Detection and nuScenes based models.
• The package can be used for deployment of the models. For deployment it is not necessary to download the dataset, it will directly work with the input pointcloud.

Content

• OpenPCDet Installation
• Installation
• Usage
• Parameters
• License
• Citation

OpenPCDet Installation

Quick Installation compatible with the setup specified above:

git clone https://github.com/open-mmlab/OpenPCDet
python3 -m pip install torch torchvision torchaudio
python3 -m pip install tensorflow
python3 -m pip install spconv-cu117
cd OpenPCDet
python3 -m pip install -r requirements.txt
python3 setup.py develop
python3 -m pip install kornia open3d
python3 -m pip install pyquaternion

Please refer to OpenPCDet’s INSTALL.md for the original installations instructions of OpenPCDet. Try the Quick Demo to verify installations.

• This project provides a docker installation reference DOCKER_INSTALL.md which is highly recommended.

Installation

Build the Packages

# GOTO the ros 2 workspace
cd src/
git clone https://github.com/Box-Robotics/ros2_numpy -b humble
python3 -m pip install catkin_pkg
sudo apt install ros-humble-ament-cmake-nose -y
python3 -m pip install nose
python3 -m pip install transform3d
git clone https://github.com/pradhanshrijal/pcdet_ros2
cd ..
rosdep install -i --from-path src --rosdistro humble -y
colcon build --symlink-install --packages-select ros2_numpy pcdet_ros2
# Source the workspace

Download the weights
Example:

• Download the PV-RCNN pre-trained weight for the Kitti Dataset [model-50M].
• Copy the weight to the checkpoints folder of the pcdet_ros2 package.

Check OpenPCDet’s Model Zoo for available models and weights.

P.S. Symlink connections will save a lot of storage space and help organise the files.

Usage

ros2 launch pcdet_ros2 pcdet.launch.py

Parameters

License

pcdet_ros2 is released under the MIT License.

Citation

If you find this project useful in your research, please consider citing the original work:

@misc{openpcdet2020,
    title={OpenPCDet: An Open-source Toolbox for 3D Object Detection from Point Clouds},
    author={OpenPCDet Development Team},
    howpublished = {\url{https://github.com/open-mmlab/OpenPCDet}},
    year={2020}
}