utbots_vision

This stack contains vision related packages, such as:

• mediapipe_track
• utbots_face_recognition
• vision_tools
• yolov8ros

And is dependant on:

• utbots_dependencies

Installation

cd ~/catkin_ws/src
git clone https://github.com/UtBotsAtHome-UTFPR/utbots_dependencies.git
git clone --recurse-submodules https://github.com/UtBotsAtHome-UTFPR/utbots_vision.git
cd ../

Dependencies

See the dependencies installation procedure for each package accessing its README.md file or, in some cases, below, in Packages Description.

Building

cd ~/catkin_ws
catkin_make -DCMAKE_BUILD_TYPE=Release
# Install ROS dependencies
sudo rosdep init
rosdep update
rosdep install --from-paths src --ignore-src -r -y

Updating

To push changes to the submodule packages (mediapipe_track, utbots_face_recognition) you should go to their repository path and perform a simple add, commit and push. After, you have to push the changes to the stack, going back to the stack repository path and doing the following command:

git submodule update --remote --merge

And then, perform a simple add, commit and push in the stack repository.

Running

See the usage explanation accessing each package in each package README.md or, in some cases, below, in Packages Description.

Packages Description

freenect_stack

• Kinect V1 Driver for ROS1. Outputs RGB sensor_msgs/Image and depth sensor_msgs/Image messages.
• Most used topics are camera/rgb/image_raw, camera/rgb/image_color and camera/depth_registered/image_raw, but several other are available.

Installation

Installed already with utbots_vision clone and pull --recurse-submodules. Can be installed separately with:

sudo apt update
sudo apt install ros-<distro>-freenect-launch

Dependencies

You must install the freenect library beforehand:

sudo apt update
sudo apt install libfreenect-dev

Running

To launch the driver and its topics, run:

roslaunch freenect_launch freenect.launch

utbots_face_recognition

Loading everything for the first time may take several seconds and subsequent launches make take some seconds if the computer is bad, have patience. For a significant number of images use cuda

Dependencies

This package is dependant on a series on problematic libraries combinations

Select a path (full path, no shortcuts) for the venv and set setup.cfg path to point to it.

pip uninstall cv_bridge # just in case so it can pull from sudo apt install ros-humble-cv-bridge
pip install deepface
pip install tf-keras
pip install --no-cache-dir --upgrade --ignore-installed opencv-python
pip install numpy==1.26.4 # This will say sucessfully installed numpy 2.2.1. Don't ask me but it works

Running

ros2 run usb_cam usb_cam_node_exe
ros2 run utbots_face_recognition recognize