Experimenting with Waterproof Connectors #3 OpenVINO in LettaPanda

OpenVINO in LettaPanda

1. Start LettaPanda

Letta Panda is special SBC supporting Windows 10, with Intel Atom Core CPU, it is definitely worth trying OpenVINO by Intel on this dev board.

First thing for SBC is to make one fully protected case against unattended damage. Totally hand work drilling holes and cutting opennings.

With HDMI display port plugged, and follow theoperation manual, just like normal all-in-one destop computer, press the ON/OFF switch, it goes as easy as anyone can do.

Refer to https://docs.lattepanda.com/content/1st_edition/power_on/

From setting-About, here is the parameters. It is typical laptop computer configuration 6~10 years , yet strong enough for AI engine.

2. Prepare OpenVINO development

Refer to https://docs.openvino.ai/latest/ for everything needed. OpenVINO is one of versitile AI toolset support Tensorflow, pytorch, paddlepaddle and OMNX.

Runtime is lightweight package and Developmen is full-functional package.In lettaPanda, development package can be installed easily with lines one by one,

python -m venv openvino_env

source openvino_env/bin/activate

python -m pip install --upgrade pip

pip install openvino-dev

Bingo, everything is ready. Only need some time for files downloading.

3. Installation of OpenVINO Notebooks

Not sure how to start, that is OK. Install OpenVINO NoteBooks and learn OpenVINO very quick.

Prerequsiton ,

• Install Python software (3.7, 3.8, 3.9, 3.10) from python.org,
• Install GIT 64 bit version of GIT from git-scm.org
• Install C++ Redistributable (option, For Python 3.8 only)

Then, activate virtual environment again

openvino_env\Scripts\activate

Do the following,

git clone --depth=1 https://github.com/openvinotoolkit/openvino_notebooks.git
cd openvino_notebooks
pip install -r requirements.txt
python -m ipykernel install --user --name openvino_env

Then jupyter notebook is ready to run with jupyter or jupyter lab,  like

jupyter 201-vision-monodepth.ipynb

4. Get to know Object Detection

The 401 notebook is WebCam SSD object detection, get to find the tennis now, importing packages first

import collections
import sys
import tarfile
import time
from pathlib import Path

import cv2
import numpy as np
from IPython import display
from openvino import runtime as ov
from openvino.tools.mo.front import tf as ov_tf_front

sys.path.append("../utils")
import notebook_utils as utils

Download the Model,

or

Transfer to IR format for OpenVINO

or

Load the Model

Now run the object detection, for one video clip, the dog is detected with box outline the object.

Now find tennis with USB video camera

That is, the start point of Underwater Camera project. just run Object detection on Letta Panda Dev Board. That is sure one easy task as the Windows supporting OpenVINO well enough for the task.

5. Next what?

Another part for lettaPanda, control periperals with Arduino Core, the C++ binding tool shall be coding to bridge the Python Code with Arduino Code. So, install Visual Studio 2017

And select C#

VS 2017 studio can be installed on LettaPanda , but it is more preferable to be installed on third hosting computer, build .exe to run on lettapanda. That is next step.