Easy face recognition try with Jetson Nano and webcam

Introduction

I tried real-time face recognition with a webcam on Jetson Nano.

• Jetson Nano
• Python3.6.8 --OpenCV3.3.1 (I intended to install 4.1.0 ...) --Webcam [(Logitech guy)](https://www.amazon.co.jp/ Logitech-Webcam-C270n-Streaming-2 year manufacturer's warranty / dp / B07QMKND9M / ref = zg_bs_2152019051_1? _Encoding = UTF8 & psc = 1 & refRID = T5TFAX38Y1CTQZ0BBXZM)

Referenced page

-How to install OpenCV -Easy-to-understand explanation about face recognition of OpenCV -How to get video with OpenCV

Image acquisition try

First, let's check how to get the image with the webcam. It was really easy.

camTest.py

import cv2

#Launch camera
capture = cv2.VideoCapture(0)

while(True):
    #Get 1 frame image
    ret, frame = capture.read()
    #Display image in window
    cv2.imshow("frame", frame)
    #Stop when q is pressed
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break
#release
capture.release()
cv2.destroyAllWindows()

It stops when I press q, but it doesn't work properly unless I press it with the focus on the window instead of the terminal.

Face recognition

When you install OpenCV, the file of the cascade classifier is included, so use it. (https://github.com/opencv/opencv/tree/master/data/haarcascades)

The procedure for face recognition is as follows

1. Get the image
2. Convert to grayscale
3. Face recognition processing
4. Write a frame at the recognition position

find_face.py

# -*- coding: utf-8 -*-

import time
import cv2

#Frame size (larger makes processing heavier)
FRAME_W = 320
FRAME_H = 240

#Cascade classifier for face detection (likely a file that summarizes features)
#I got an error when I tried to read something that was in another folder, so I copied it to the same folder
cascadeFilePath = './haarcascade_frontalface_default.xml'
cascade = cv2.CascadeClassifier(cascadeFilePath)

#Camera settings
cam = cv2.VideoCapture(0)
time.sleep(1)                 #Waiting for startup (for the time being)
cam.set(cv2.CAP_PROP_FPS, 60) #I don't know if it was 60
cam.set(cv2.CAP_PROP_FRAME_WIDTH, FRAME_W)
cam.set(cv2.CAP_PROP_FRAME_HEIGHT, FRAME_H)

while(True):
    #End with q
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

    #Image acquisition
    ret, frame = cam.read()
    #Convert to grayscale
    gray_image = cv2.cvtColor(frame, cv2.COLOR_RGB2GRAY)
    #Face recognition
    facerect = cascade.detectMultiScale(gray_image, scaleFactor=1.1, minNeighbors=2, minSize=(30, 30))

    #Was the face detected?
    if len(facerect) > 0:
        #Border color
        line_color = (255, 102, 51)
        #Text color
        font_color = (255, 102, 51)

        #Write a frame and FACE characters on the detected face
        for (x, y, width, height) in facerect:
            cv2.rectangle(frame, (x, y), (x + width, y + height), line_color, 2)
            cv2.putText(frame, 'FACE', (x, y), cv2.FONT_HERSHEY_SIMPLEX, 0.7, font_color, 1, cv2.LINE_AA)

    #Show in window
    cv2.imshow('frame', frame)
    

#End processing
cam.release()
cv2.destroyAllWindows()

Recognized in real time!