Body and hand detection: I apply the lightweight Openpose model, MediaPipe library, and the optical flow algorithm. The combination can process in real-time and with a lightweight model, so it is suitable for edge devices.
As any other component, BodyHandJointsDetector needs a configuration file to start. In
etc/config
The configuration is already set, if changing the port of this component, please change to the same port in this component's client link
The setting for python enviroment can be found here
Copy pretrained model from this link to src/_model/ in the detector component folder.
To avoid changing the config file in the repository, we can copy it to the component's home directory, so changes will remain untouched by future git pulls:
cd <BodyHandJointsDetector's path>
cp etc/config config
Need to CMAKE the code with this command:
cmake .
After editing the new config file we can run the component:
python src/BodyHandJointsDetector.py etc/config