Real-time Face Recognition Project with Raspberry Pi and Intel Neural Compute Stick 2

Real-time Face Recognition Project with Raspberry Pi and Intel Neural Compute Stick 2

by

If you don’t want to miss my updates, remember to check the public account in the upper right corner and set it as a star, take down the star and give it to me!

Real-time Face Recognition Project with Raspberry Pi and Intel Neural Compute Stick 2
Real-time Face Recognition Project with Raspberry Pi and Intel Neural Compute Stick 2

I am Tongji Zihai, when I received the invitation from Darlwen to do the Raspberry Pi project, I happened to be in the United States and had the privilege to go to Intel headquarters to learn more about NCS2, so I decided to combine the two to create an artificial intelligence application.

At Intel headquarters
After returning to China, I dug out my long-hidden Raspberry Pi 3B+ (of course, the latest Pi4 is also completely fine) and prepared to play with it.For those who have not been in contact with Raspberry Pi, you can think of it as a mini-computer with a Linux system. It looks very small and exquisite, but its functions are quite powerful.The Raspberry Pi comes pre-installed with a Linux system, is small in size, equipped with an ARM architecture processor, and has various interfaces, so it can do everything that a computer can do.
I have previously done some small projects with Raspberry Pi, basically using OpenCV to complete them.As a result, I found that the Raspberry Pi could not achieve real-time tracking of computer vision, and it was quite strenuous to run.The Intel Neural Compute Stick 2 (NCS2) can support the Raspberry Pi very well, so can it run networks on the Raspberry Pi?This was something I never dared to imagine before.After getting both things, I decided to do an experiment:

Summarizing some key operation steps for everyone, if you want to learn step by step completely, click to read the original text for more details.

Materials Preparation

– Raspberry Pi 3B+ motherboard
– Intel Movidius Neural Compute Stick 2 (Intel® Neural Compute Stick 2)
– Camera
– Display (not necessarily attached to the board, can also connect to a common display)

– A computer

Configuring the Raspberry Pi System

Configuration path:Download image -> Burn image -> Configure Raspberry Pi
To download the image, you need to change the default download source of Raspberry Pi from abroad to Tsinghua University’s open-source software mirror site.
In the command line interface of the Raspberry Pi, enter:
sudo nano /etc/apt/sources.list
Use the keyboard arrow keys to control, add a # at the beginning of the first line, and copy the following content to the last line, as shown in the image:
deb http://mirrors.tuna.tsinghua.edu.cn/raspbian/raspbian/ stretch main contrib non-free rpi deb-src http://mirrors.tuna.tsinghua.edu.cn/raspbian/raspbian/ stretch main contrib non-free rpi

Real-time Face Recognition Project with Raspberry Pi and Intel Neural Compute Stick 2

First press ctrl+o on the keyboard, then press enter to save, and then press ctrl+x to exit the nano editor and return to the command line interface. Then enter the following command to update to the latest software list from Tsinghua University’s mirror source:
sudo apt-get update
This command will access each URL in the source list, read the software list, and then save it locally on the Raspberry Pi.
The above steps have achieved changing the source of the Raspbian package manager apt-get to Tsinghua University’s software mirror site and updating the software list. In the future, when executing sudo apt-get install software_name in the Raspberry Pi command line, it will automatically download from Tsinghua University’s open-source software mirror site at high speed.
After the source file is downloaded, the next step is to configure the internal settings of the Raspberry Pi, install commonly used software, etc., the detailed steps can be found by replying:Raspberry Pi.

This step is the necessary path summarized by countless Raspberry Pi players over the years, condensing the blood, tears, and sweat of many predecessors, especially suitable for beginners to read.

Check

uname -m

Output armv7l indicates normal operation.

Change Source

sudo nano /etc/apt/sources.list
Comment out the original file content with # and replace it with the following content:
deb http://mirrors.tuna.tsinghua.edu.cn/raspbian/raspbian/ stretch main contrib non-free rpi deb-src http://mirrors.tuna.tsinghua.edu.cn/raspbian/raspbian/ stretch main contrib non-free rpi
Save and exit.
sudo nano /etc/apt/sources.list.d/raspi.list
Comment out the original file content with # and replace it with the following content:
deb http://mirror.tuna.tsinghua.edu.cn/raspberrypi/ stretch main ui deb-src http://mirror.tuna.tsinghua.edu.cn/raspberrypi/ stretch main ui

Save and exit. Use the sudo apt-get update command to update the software source list.

Install CMake

sudo apt install cmake

Download OpenVINO Toolkit for Raspbian

Use your computer to click the link below to download the latest version of OpenVINO released on October 25, 2019, and then use Filezilla to transfer it to the Downloads folder of the Raspberry Pi:
https://download.01.org/opencv/2019/openvinotoolkit/R3/l_openvino_toolkit_runtime_raspbian_p_2019.3.334.tgz
cd ~/Downloads/tar -xf l_openvino_toolkit_runtime_raspbian_p_2019.3.334.tgz
Rename the extracted files to: inference_engine_vpu_arm
source inference_engine_vpu_arm/bin/setupvars.sh sh inference_engine_vpu_arm/install_dependencies/install_NCS_udev_rules.sh sed -i "s|<INSTALLDIR>|$(pwd)/inference_engine_vpu_arm|" inference_engine_vpu_arm/bin/setupvars.sh

Set Environment Variables

sudo nano /home/pi/.bashrc
Add the following content at the end:
source /home/pi/Downloads/inference_engine_vpu_arm/bin/setupvars.sh
Real-time Face Recognition Project with Raspberry Pi and Intel Neural Compute Stick 2
Open a new command line window, and if the prompt shows:[setupvars.sh] OpenVINO environment initialized, it indicates the configuration was successful.

Configure USB Rules for Intel Neural Stick 2

sudo usermod -a -G users "$(whoami)" sh inference_engine_vpu_arm/install_dependencies/install_NCS_udev_rules.sh
Tip:
Updating udev rules... Udev rules have been successfully installed.
Indicates that the installation was successful.
Insert the Intel Neural Stick into the USB port of the Raspberry Pi and compile:
cd inference_engine_vpu_arm/deployment_tools/inference_engine/samples mkdir build && cd build cmake .. -DCMAKE_BUILD_TYPE=Release -DCMAKE_CXX_FLAGS="-march=armv7-a" make -j2 object_detection_sample_ssd

Download Model Files and Weights for Neural Stick

wget --no-check-certificate https://download.01.org/openvinotoolkit/2018_R4/open_model_zoo/face-detection-adas-0001/FP16/face-detection-adas-0001.bin wget --no-check-certificate https://download.01.org/openvinotoolkit/2018_R4/open_model_zoo/face-detection-adas-0001/FP16/face-detection-adas-0001.xml

Face Detection

Replace <path_to_image> with the absolute path of your face image

pi@raspberrypi:~/Downloads/inference_engine_vpu_arm/deployment_tools/inference_engine/samples/build$ ./armv7l/Release/object_detection_sample_ssd -m face-detection-adas-0001.xml -d MYRIAD -i /home/pi/Downloads/face/1.jpg

Call OpenCV for Face Detection in Python

import cv2 as cv print('Starting face detection') # Load model and weight files net = cv.dnn.readNet('face-detection-adas-0001.xml', 'face-detection-adas-0001.bin') # Specify target device net.setPreferableTarget(cv.dnn.DNN_TARGET_MYRIAD) # Read image, change the image path to your face image frame = cv.imread('./6.jpg') # Prepare input blob and perform an inference blob = cv.dnn.blobFromImage(frame, size=(672, 384), ddepth=cv.CV_8U) net.setInput(blob) out = net.forward() # Draw detected faces on the frame for detection in out.reshape(-1, 7): confidence = float(detection[2]) # xmin,ymin - top left corner coordinates xmin = int(detection[3] * frame.shape[1]) ymin = int(detection[4] * frame.shape[0]) # xmax,ymax - bottom right corner coordinates xmax = int(detection[5] * frame.shape[1]) ymax = int(detection[6] * frame.shape[0]) if confidence > 0.5: cv.rectangle(frame, (xmin, ymin), (xmax, ymax), color=(0, 255, 0)) # Save the detected face image cv.imwrite('./out.png', frame) print('Face detection completed, saved the result image')

Configure the Raspberry Pi Camera

sudo apt-get install python-opencv sudo apt-get install fswebcam
Insert the Raspberry Pi camera and configure the Raspberry Pi camera
sudo nano /etc/modules
Add a line at the end of this file
bcm2835-v4l2

Restart Raspberry Pi

Enter the command:
vcgencmd get_camera
If you get the following result, it proves that the camera is connected successfully:
raspistill -o image.jpg
Call the camera to take a photo named image.jpg, stored in /pi/home path, which is the path displayed when you open the file explorer in the upper left corner. If you see the red light on the camera and the directory contains a photo, it further indicates that the camera is configured correctly.
ls /dev

Real-time Face Recognition Project with Raspberry Pi and Intel Neural Compute Stick 2

The video0 in the lower right corner is the camera.
fswebcam 10 test.jpg

After execution, it will delay 10 frames to take a picture (give some preparation time) and generate an image named test, stored in the /home/pi directory.

Configure Display Window

Open VNC Viewer, run this command in the Raspberry Pi command line

export DISPLAY=:0.0

Real-time Face Detection with Raspberry Pi Camera

# coding=utf-8 # face-detection-camera.py import cv2 as cv import numpy as np print('Starting real-time face detection with camera') # Load model files and weights net = cv.dnn.readNet('face-detection-adas-0001.xml','face-detection-adas-0001.bin') # Specify target device net.setPreferableTarget(cv.dnn.DNN_TARGET_MYRIAD) # Read image frames from camera cap=cv.VideoCapture(0) while(1): # Get a frame image ret,frame=cap.read() # Prepare input blob and perform an inference frame=cv.resize(frame,(480,320),interpolation=cv.INTER_CUBIC) blob = cv.dnn.blobFromImage(frame, size=(672, 384), ddepth=cv.CV_8U) net.setInput(blob) out = net.forward() # Draw face boxes for detection in out.reshape(-1, 7): confidence = float(detection[2]) # Get the coordinates of the top left corner of the image xmin = int(detection[3] * frame.shape[1]) ymin = int(detection[4] * frame.shape[0]) # Get the coordinates of the bottom right corner of the image xmax = int(detection[5] * frame.shape[1]) ymax = int(detection[6] * frame.shape[0]) if confidence > 0.5: cv.rectangle(frame, (xmin, ymin), (xmax, ymax), color=(0, 255, 0)) # Display the image cv.imshow("capture",frame) if cv.waitKey(1) & 0xFF == ord('q'): # Listen for keyboard actions every 1 millisecond, press q to end and save the image cv.imwrite('out.png', frame) print("save one image done!") break # Close the camera and display window cap.release() cv.destroyAllWindows() print('Real-time face detection with camera completed')

Run this script file

python3 face-detection-camera.py
You can see the real-time recognized face image on the display:

Real-time Face Recognition Project with Raspberry Pi and Intel Neural Compute Stick 2

With NCS2, the Raspberry Pi can run the computer algorithm applications you want with ease. Of course, the premise is that you have learned the most basic applications. Below is the detailed course prepared by Zihai for beginners, you canclick to read the original textfor more details.

Real-time Face Recognition Project with Raspberry Pi and Intel Neural Compute Stick 2

END
Real-time Face Recognition Project with Raspberry Pi and Intel Neural Compute Stick 2
Darlwenaccompanied by the globally renowned component distributorDigi-Keyin the most important and final lesson of your university career:Graduation Project!

The graduation project course consists of 17 lectures: from “topic selection – supplementing basic knowledge points – project breakdown explanation – excellent work sharing” to help you solve various problems from the topic proposal to the final defense process.

Dejie – 17 free graduation design courses high-energy review:

Introduction: Graduation project “pass” secrets, attached board card white piao

Graduation project lesson 1: How to choose a topic, which type of graduation project is suitable for you

Graduation project lesson 2:Graduation project reading case analysis

Graduation project lesson 3:Graduation project development platform selection

Graduation project lesson 4/5:Use AD19 to draw STM32 core board in 2 hours

Graduation project lesson 6:C language/Verilog language basics supplement

Graduation project lesson 7: Build a neural network with Tensorflow (read the original text for more details)

Graduation project lesson 8: System debugging skills and instrument usage

Graduation project lesson 9:Use instruments to speed up design and development

Graduation project lesson 10:Power circuit design

Graduation project lesson 11:Analysis and design of commonly used sensors

Graduation project lesson 12:Analysis of commonly used data buses (read the original text for more details)

Graduation project lesson 13:Communication interface design

Graduation project lesson 14:Teach you to design motor drive circuits

Graduation project lesson 15:Open source STM32 IoT smart home project

Real-time Face Recognition Project with Raspberry Pi and Intel Neural Compute Stick 2

Real-time Face Recognition Project with Raspberry Pi and Intel Neural Compute Stick 2
Click the original text to learn the course!

Leave a Comment