Energirobotter Vision

Vision capabilities and camera handling for the Humanoid Robot project "Energirobotter".

Setup

Clone this repository into a workspace/src/ folder, along with realsense-ros (run commands from root of workspace):

git clone [email protected]:energinet-digitalisering/energirobotter-vision.git src/energirobotter-vision/
git clone https://github.com/IntelRealSense/realsense-ros.git src/realsense-ros/

Setup RealSense SDK:

sudo apt install ros-humble-librealsense2*

Dependencies

In worspace root, source ROS and install ROS dependencies with rosdep:

source /opt/ros/humble/setup.bash
rosdep install -i --from-path src --rosdistro $ROS_DISTRO --skip-keys=librealsense2 -y

Python modules not included in rosdistro can be installed from root of workspace with:

pip install -r src/energirobotter-vision/requirements.txt

AI model

Download face detection model yolov8n-face.pt from the yolo-face repository. Move the model into the src/energirobotter-vision/face_detection/models/ directory.

Build

Build workspace with:

colcon build --symlink-install

Usage

Use the vision_bringup package's vision.launch.py to start the camera and face detection:

source install/setup.bash
ros2 launch vision_bringup vision.launch.py use_compressed:=true

If no camera is available, run example with:

ros2 launch vision_bringup vision.launch.py use_mock_camera:=true

Note: The first time face_detection is run ultralytics will install som extra packages, and require a restart of the node.