OpenVINO Toolkit for Raspbian OS

Some scripts to install OpenVINO Toolkit on Raspberry Pi.

All resources including OpenVINO Toolkit and Open Model Zoo are downloaded from official web-site: https://docs.openvinotoolkit.org/.

Given scripts succesfully tested with:

• Rapberry Pi 3 B+;
• OS Raspbian Buster with desktop by 2020.20 (4.19);
• Intel Neural Compute Stick 2;
• Raspberry Pi Camera Module v2.

step 1

Run script install_openvino.sh:

bash install_openvino.sh

Whis script will update and upgrade your Raspbian OS, download and unpack archive with OpenVINO Toolkit v2020.1, update .bashrc, install some dependencies and clone repository with demo code.

step 2

Download models required with demo you would like to run. I found some troubles with compability between available OpenVINO Toolkit version and Open Model Zoo. So that's why I recommend to use 2019.R3 models with 2020.1 toolkit. Experiments showed that everything works fine. To download models use:

bash download_models.sh <mode>

where mode is name of demo model (like *.py file). For example: face_recognition_demo. Available models:

• face_recognition_demo.

Models will be downloaded to ./models.

step 3

Now python virtual environment need to be created. For this porpose just run script:

bash setup_python_env.sh

This script will create virtual environment in ~/venvs/openvino and install python-packages from requirements.txt. Later please use this environment to run demo code from Open Model Zoo.

step 4

Run your model with

bash demo.sh <model>

where mode is name of demo model. For example: face_recognition_demo. Available models:

• face_recognition_demo.

Be sure you downloded required models before (check step 2). I recommend to run demos via VNC (remote control with desktop). It will allow to see inference output in realtime. Also it's assumed that you use Neural Compute Stick 2 and Raspberry Camera connected to your Raspberry Pi. If you would like to change some arguments to run some models please check and update demo.sh.

Some notes about running models:

• face_recognition_demo:
  • be sure you put photos of your faces in ./resources/face_recognition_demo/faces in format <name>-<order-num>.jpg, where <name> is name of person (for example Jack or Julia) and <order-num> is number from 0 to inf.

---

Write to me if you have any queistions! Good luck and have a nice day!