Real-time Object Detection with Raspberry Pi and Intel NCS 2

Inspiration

I recently received an Intel Neural Compute Stick 2 from a deep learning challenge organized by Intel. This weekend I decided to try out this VPU (Vision Processing Unit) device with my Raspberry pi 4.

Hardware and Software Used

Raspberry Pi 4 Raspberry Pi camera Intel Neural Compute Stick 2 Intel OpenVINO toolkit OpenCV 3 Python 3

What it does

The system can detect some of the everyday objects like person, cats, dogs, cows, birds, bottle, airplane. It uses the MobileNet SSD object detection model.

How I built it

I followed the official documentation for downloading the Intel OpenVINO toolkit on my Raspberry Pi 4 OS and also to set up my Intel NCS2 with the Raspberry Pi 4.

Link for installation steps of OpenVINO on Raspberry Pi : link

Link for setting up my Intel NCS2 with raspberry Pi 4 : link

Main motivation by pyimagesearch blog : link

Accomplishments that I'm proud of

My model can successfully detect the following objects: dog, cow, cat, bird, person, bottle , tv monitor, chair, bus, car.

What I learned

This weekend, I got to get my hands on a new device, i.e. , the Intel Neural Compute Stick 2 and also I got to work a lot with the dnn module of OpenCV in Python. I also tried out the Intel OpenVINO toolkit on the Raspberry Pi 4.

What's next for Real-time Object Detection in Raspberry Pi

For next step for this project, I would like to further improve the code and try out some new features. I plan to implement Tiny-YOLO algorithm as a next step.

Built With

• camera
• intel
• intelnncs2
• mobilenetssd
• ncs2
• objectdetection
• opencv
• openvino
• python
• raspberry-pi
• raspberrypi