提供了一个基于Ultralytics YOLOv8的ROS功能包。该功能包已在Ubuntu 18.04上进行了测试。
- Ubuntu 18.04
- ROS melodic
- Python3.8与2.7环境，PyTorch == 1.10.0(Jetson TX2安装方法, Pytorch源代码, torchvision)
- 克隆代码及其子模块
git clone https://gitee.com/lisq58/excavator_detect.git
git submodule update --init --recursive
-
检查 src/mavlink, src/mavros, src/geometry2, src/vicon_bridge目录下子模块均有内容即可
-
安装mavros的依赖与地理列表数据集
sudo apt-get install ros-melodic-geographic-msgs ros-melodic-mavros ros-melodic-mavros-extras libgeographic-dev -y
sudo ./src/mavros/mavros/scripts/install_geographiclib_datasets.sh
- 安装realsense的依赖
sudo apt-get install ros-melodic-realsense2-camera ros-melodic-realsense2-description -y
- 修改
~/.bashrc使得其他节点能调用源代码编译的tf2
echo "export PYTHONPATH=$PYTHONPATH:~/workspace/excavator_detect/src" >> ~/.bashrc
source ~/.bashrc
- 安装python3.8
sudo apt-get install python3.8 python3-pip -y
- 创建python3.8虚拟环境
sudo -H python3.8 -m pip install virtualenv virtualenvwrapper
sudo rm -rf ~/.cache/pip
echo "export WORKON_HOME=$HOME/.virtualenvs" >> ~/.bashrc
echo "export VIRTUALENVWRAPPER_PYTHON=/usr/bin/python3" >> ~/.bashrc
source ~/.bashrc
- 使用virtualenv创建虚拟环境，virtualenvname为虚拟环境名称，-p表示母python路径，如/usr/bin/python3.8 , 创建成功后启动虚拟环境
virtualenv path/to/virtualenvname -p path/to/python3
source path/to/vitualenvname/bin/activate
- 在虚拟环境中安装依赖, 若权重文件使用.onnx类型则额外安装onnx onnxruntime
pip install ultralytics pyrealsense2 apriltag rospkg
- 安装编译ROS包的依赖
sudo apt-get install python-catkin-tools libxml2 libxml2-dev libxslt1.1 libxslt1-dev -y
- 清除以往的编译文件
catkin clean
- 安装编译的依赖(非虚拟环境)
python3.6 -m pip install empy==3.3.4 catkin_pkg futur lxml
sudo apt-get install python-pip
python2 -m pip install empy==3.3.4
- 使用python3解释器编译在python3环境下运行的节点(不要在虚拟环境编译)
catkin build mavlink mavros geometry2 cv_joint_angle --cmake-args -DCMAKE_BUILD_TYPE=Release -DPYTHON_EXECUTABLE=/usr/bin/python3 -DPYTHON_INCLUDE_DIR=/usr/include/python3.6m -DPYTHON_LIBRARY=/usr/lib/x86_64-linux-gnu/libpython3.6m.so
- 编译其余节点
catkin build
- 将
/opt/ros/melodic/share/cv_bridge/cmake/文件94行最后一个索引改为/usr/include/opencv4
sudo gedit /opt/ros/melodic/share/cv_bridge/cmake/cv_bridgeConfig.cmake
- 首先，确保将训练好的权重放在 cv_joint_angle/weights 文件夹中。
- yolov8.launch文件 中设置权重文件与推理相关参数设置
- 将环境变量加入bash
echo "source your/workspace/devel/setup.bash" >> ~/.bashrc
source ~/.bashrc
- 使用该功能包
./your/workspace/excavator_detect/run.sh