OpenVINO people tracker

Simple PoC capable of detect people in a scene taken by cam or by a previously recordered video.
Mainly used to test the succesful installation of OpenVINO toolkit.

Install & run

Python3 and virtualenv packages must be installed on your system.

virtualenv venv
source venv/bin/activate
pip install -r requirements.txt
python main.py <device> [source device]

This will execute the pople counter application on your PC.
The <device> argument must be replace with CPU or GPU, depending on which device you want to perform the inference.

By default the /dev/video0 stream will be used. If you want to use a different source, let's specify it as a second argument.

Docker image

You can use the dockerdima/clea_ai_base docker image to test the application or to run other applications.
To do that, create a container with the following command:

docker run --rm -it --privileged --net=host -v /tmp/.X11-unix:/tmp/.X11-unix \
            -e DISPLAY=$DISPLAY --device /dev/dri --device /dev/video0:/dev/video0 \
            dockerdima/clea_ai_base

and type

cd ws/people-tracker-OpenVINO/
source venv/bin/activate
python main.py GPU

In case of problems related to X11 server, type
xhost +
on a different console to accept incoming connections from any client.