Usage

The nodes in this repo are run in a python3 virtual environment. As roslaunch interferes with this, the nodes are launched as python scripts

Installation

• Create a Python3 virtualenv as an overlay of the ROS Noetic Python interpreter (in order to have access to rospy components). In the following it assumed that the path to the virtualenv folder is $BOX_YOLO_ENV:

  export BOX_YOLO_ENV=~/.virtualenvs/box_yolo;
  mkdir -p ${BOX_YOLO_ENV};
  python3 -m venv --system-site-packages ${BOX_YOLO_ENV};
  source ${BOX_YOLO_ENV}/bin/activate;

• In the following, it is assumed that you have created a ROS Noetic catkin workspace called $BOX_CATKIN_WS, in which this package should be installed.

  source /opt/ros/noetic/setup.bash
  source ${BOX_CATKIN_WS}/devel/setup.bash
  cd ${BOX_CATKIN_WS}/src;
  # Clone this repo.
  git clone [email protected]:harmony-eu/box_yolo_detection.git;
  cd ${BOX_CATKIN_WS}/src/box_yolo_detection;
  # Clones the submodule of this repo (the original `yolov5` repo).
  git submodule update -i -r;
  pip3 install -r requirements.txt
  # Installs the required PyTorch dependencies (tested with the versions below).
  pip3 install torch torchvision --extra-index-url https://download.pytorch.org/whl/cu116

• Put your model files (.pt and .yaml) in the models folder.
• To run the nodes in this repo, you will need to have built ROS messages of type detection_msgs. To do so, clone the detection_msgs repo in your catkin workspace and build it:

  cd ${BOX_CATKIN_WS}/src;
  git clone https://github.com/mats-robotics/detection_msgs.git;
  catkin build