OpenCV

1. What is OpenCV ?
Open CV is an open source library for image processing . The application includes, face detection, objection tracking, land mark detection and more

2. What language does OpenCV use ?
Open CV uses multiple language such as C++ , phyton, java. However, it is preferable to use python. 

3. Prerequisite of OpenCV ?
a. Read and write image using OpenCV
b. Diffrerentiate color spaces present in OpenCV
c. Create NumPy array and familiar with the basic operation of NumPy

Pre-OpenCV 

1. Install the latest Python Software
Link: https://www.python.org/downloads/
(Note: At this time of learning, I'm using the 3.11 version)
2. Install Visual Studio Code 
Link: https://code.visualstudio.com/download 
3. Search "Command Prompt" on the device and install two packages
a. opencv-python module by typing the following in Command Prompt
pip install opencv-contrib-python
b. caer by typing the following in Command Prompt
pip install caer

Module 

Read images 
1. Create a folder in Document - "OpenCV"
Create two folders such as Videos and Photos and download few photos and videos in the Document folder
2. Go to Visual Code Studio 
Open the OpenCV folder
Create a new file - read.py 
3. Code 

import cv2 as cv

img = cv.imread('Photos/cat.jpg) // this reads the image

cv.imshow('cat',img) // this display the image 

cv.waitKey(0)

4. Run the code . Make sure Command prompt is used in the Terminal

Then type python read.py and the image will appear 

Read Video

Code 

import cv2 as cv

# img = cv.imread('Photos/cat1.jpeg')

# cv.imshow('Cat',img)

capture = cv.VideoCapture('Videos/Test Flight.mp4') # capture is variable . inside () - 1. a path to video file as the current code 2. as numbers can be assigned or inserted. Example : (1) - Camera/Webcam connected to laptop

while True:

    isTrue,frame = capture.read()# read video frame by frame

    cv.imshow('Video',frame) # display as individual frame

    if cv.waitKey(20) & 0xFF==ord('d'): #stop the video from play indefinately when d is pressed

        break # break while loop

capture.release ()

cv.destroyAllWindows()

2. Run the code
(Note : Press "d" to stop the video)

Resizing and Rescaling Image

import cv2 as cv

img = cv.imread('Photos/cat1.jpeg')

cv.imshow('Cat',img)

def rescaleFrame(frame, scale =0.75) :

     # create a function . 0.75 - resized to 75 percent

     # works for images,videos and live videos

    width = int (frame.shape[1] * scale) #  0 is width

    height = int (frame.shape[0] * scale) #  1 is height

    dimensions = (width,height)

    return cv.resize(frame,dimensions,interpolation = cv.INTER_AREA)

resized_image =rescaleFrame(img)

cv.imshow('Image',resized_image)

# def changeRes(width,height):

    #only for live videos

    #capture.set(3,width)

    #capture.set(4,height)

#read videos

capture = cv.VideoCapture('Videos/Test Flight.mp4') # capture is variable . inside () - 1. a path to video file as the current code 2. as numbers can be assigned or inserted. Example : (1) - Camera/Webcam connected to laptop

while True:

    isTrue,frame = capture.read()# read video frame by frame

    frame_resized = rescaleFrame(frame)

    cv.imshow('Video',frame) # display as individual frame

    cv.imshow('Video Resized', frame_resized)

    if cv.waitKey(20) & 0xFF==ord('d'): #stop the video from play indefinitely when d is pressed

        break # break while loop

capture.release ()

cv.destroyAllWindows()

Draw and Write on Images 

Code 1 : Display blank image

import cv2 as cv

import numpy as np

blank = np.zeros((500,500),dtype = 'uint8') #size of blank image

cv.imshow('Blank',blank)

img = cv.imread('Photos/cat1.jpeg')

cv.imshow('Cat',img)

cv.waitKey(0)

Code 2 : Paint the image a certain color

import cv2 as cv

import numpy as np

blank = np.zeros((500,500,3),dtype = 'uint8')#(height,width,number of colour channels) #size of blank image

cv.imshow('Blank',blank)

blank [:] = 0,255,0 # colour . Explore how to do this

cv.imshow('Green',blank)

cv.waitKey(0)

Code 3 : Draw a rectangle 

import cv2 as cv

import numpy as np

blank = np.zeros((500,500,3),dtype = 'uint8')#(height,width,number of colour channels) #size of blank image

cv.imshow('Blank',blank)

cv.rectangle(blank,(0,0),(250,250),(0,255,0), thickness = 2)

#(0,0) - origin

#(250,250) - end point

#(0,255,0) - green

#thickness = 2 - line thickness #thickness = cv.FILLED - to fill image with green #thickness = -1 - to fill image with green

cv.imshow('Rectangle',blank)

cv.waitKey(0)

Code 4: Draw a circle

import cv2 as cv

import numpy as np

blank = np.zeros((500,500,3),dtype = 'uint8')#(height,width,number of colour channels) #size of blank image

cv.imshow('Blank',blank)

cv.circle (blank,(250,250),40,(0,0,255) , thickness = 2)

#(250,250) - centre of frame

# 40 - radius

#(0,0,255) - red

#thickness = 2 - line thickness

cv.imshow('circle',blank)

cv.waitKey(0)

Code 4 - Draw line 

import cv2 as cv

import numpy as np

blank = np.zeros((500,500,3),dtype = 'uint8')#(height,width,number of colour channels) #size of blank image

cv.imshow('Blank',blank)

cv.line(blank,(0,250),(250,250),(255,255,255) , thickness = 2)

#(0,250) - origin

#(250,250) - end

#(255,255,255) - white

cv.imshow('line',blank)

cv.waitKey(0)

Code 5 - Write text 

import cv2 as cv

import numpy as np

blank = np.zeros((500,500,3),dtype = 'uint8')#(height,width,number of colour channels) #size of blank image

cv.imshow('Blank',blank)

cv.putText(blank,'Hello', (225,225), cv.FONT_HERSHEY_SCRIPT_COMPLEX,1.0,(0,255,0), thickness=2)

# 'Hello' - text

# (255,255) - coordinate of text

# cv.FONT - font

# 1.0 - scale

# (0,255,0) - color

# thickness = 2 - line thickness

cv.imshow('Text',blank)

cv.waitKey(0)

5 Essential Function

Code 1 : Converting image to gray scale

import cv2 as cv

img = cv.imread('Photos/cat1.jpeg')

cv.imshow('cat1',img)

gray = cv.cvtColor(img,cv.COLOR_BGR2GRAY) #cv.COLOR_BGR2GRAY- convert color (from)to(another color)

cv.imshow('Gray',gray)

cv.waitKey(0)

Code 2 : Blurring image

import cv2 as cv

img = cv.imread('Photos/cat1.jpeg')

cv.imshow('cat1',img)

blur = cv.GaussianBlur(img,(3,3), cv.BORDER_DEFAULT)

# (3,3) - blurness . increase this and pic gets blurer

cv.imshow('Gray',blur)

cv.waitKey(0)

Code 3: Edge Cascade

import cv2 as cv

img = cv.imread('Photos/cat1.jpeg')

cv.imshow('cat1',img)

canny =cv.Canny(img,125,175)

# 125.175 - threshhold value . change img to blur to reduce the amount of edges

cv.imshow ('Canny edges',canny)

cv.waitKey(0)

Result 

Code 4 : Dilate image

import cv2 as cv

img = cv.imread('Photos/cat1.jpeg')

cv.imshow('cat1',img)

canny =cv.Canny(img,125,175)

# 125.175 - threshhold value

cv.imshow ('Canny edges',canny)

dilated = cv.dilate(canny,(7,7),iterations =3)

cv.imshow('cat1',dilated)

cv.waitKey(0)

Result

Code 5: Erode Image

import cv2 as cv

img = cv.imread('Photos/cat1.jpeg')

cv.imshow('cat1',img)

canny =cv.Canny(img,125,175)

# 125.175 - threshhold value

cv.imshow ('Canny edges',canny)

dilated = cv.dilate(canny,(7,7),iterations =3)

cv.imshow('cat1',dilated)

eroded = cv.erode(dilated,(7,7),iterations =3)

cv.imshow('cat1',eroded)

cv.waitKey(0)

Result 

Code 5 : Resize and Crop

import cv2 as cv

img = cv.imread('Photos/cat1.jpeg')

cv.imshow('cat1',img)

#Resize

resized = cv.resize(img,(500,500),interpolation =cv.INTER_CUBIC)

#(500,500) - destination size  or size of resizing

# cv.INTER_LINEAR - to scale to larger image

# cv.INTER_CUBIC - to scale to larger image with better quality

# cv.INTER_AREA - to shrink

cv.imshow('resized', resized)

cv.waitKey(0)

Code 6 : Crop (not verified)

import cv2 as cv

img = cv.imread('Photos/cat1.jpeg')

cv.imshow('cat1',img)

#crop using array slicing

cropped = img[50 :200, 200:400]

#[50:200,200:400] - region to crop

cv.imshow('Cropped', cropped)

cv.waitKey(0)

Translating Image 

import cv2 as cv

import numpy as np

img = cv.imread('Photos/cat1.jpeg')

cv.imshow('cat1',img)

#translation - shifting image along x -axis and y-axis

def translate(img,x,y):

    #(img,x,y) -  x and y represents number of pixels you want to shift along the x-axis and y-axis

    trasMat = np.float32([[1,0,x],[0,1,y]])

    #transMat - which is known as translation matrix is used to translate image

    dimensions = (img.shape[1], img.shape[0])

    # img.shape[1] - width

    # img.shape[0]- height

    return cv.warpAffine(img,trasMat,dimensions)

# -x -- Left

# -y -- Up

# x -- right

# y --Down

translated = translate(img, 100, 100)

# (img, 100, 100) - shifting right by 100 pixels and down by 100 pixels

cv.imshow('translated',translated)

cv.waitKey(0)

Rotation

import cv2 as cv

import numpy as np

img = cv.imread('Photos/cat1.jpeg')

cv.imshow('cat1',img)

#rotation by some angle

def rotate (img,angle,rotPoint=None):

#rotation function

    (height,width) = img.shape[:2]

    if rotPoint is None:

        rotPoint = (width//2,height//2)

        #rotate around centre

    rotMat = cv.getRotationMatrix2D(rotPoint,angle,1.0)

    #rotMat - rotation matrics

    #(rotPoint,angle,1.0) - rotationpoint, angle, scale value

    dimensions = (width,height)

    return cv.warpAffine(img,rotMat,dimensions)

rotated = rotate(img,45)

cv.imshow('rotated',rotated)

rotated_rotated = rotate(rotated, -45)

#rotated_rotated = rotate(rotated,45) - rotated, rotated image

cv.imshow('rot',rotated_rotated)

cv.waitKey(0)

Fliping  image 

import cv2 as cv

import numpy as np

img = cv.imread('Photos/cat1.jpeg')

cv.imshow('cat1',img)

#flip

flip = cv.flip(img,1)

# 0 - flipping image vertically (x-axis)

# 1 - flipping image horizontall (y-axis)

# -1 -flipping image both harizontally and vertically

cv.imshow('flip',flip)

cv.waitKey(0)

Contour Detection

- to find out how many contours are there

import cv2 as cv

img = cv.imread('Photos/cat1.jpeg')

cv.imshow('Cats',img)

#contour detection - which are boundaries of object. contours and edges are two different things

gray = cv.cvtColor(img,cv.COLOR_BGR2GRAY)

#convert image to grey scale

cv.imshow('gray', gray)

canny = cv.Canny(img,125,175)

#grab the edges

# (125,175) -threshold value

cv.imshow('canny',canny)

contours,hierarchies = cv.findContours(canny,cv.RETR_LIST,cv.CHAIN_APPROX_NONE)

#cv.(something as below) - mode to find contour

#cv.RETR_TREE - all hierarchical contour

#cv.RETR_EXTERNAL - external contour only

#cv.RETR_LIST - all contour

#cv.CHAIN_APPROX_NONE - contour approximation ; take all points

#cv.CHAIN_APPROX_SIMPLE - only takes end points

print(f'{len(contours)} contour(s) found')

cv.waitKey(0)

Result 

Finding edges by blurring image

import cv2 as cv

img = cv.imread('Photos/cat1.jpeg')

cv.imshow('Cats',img)

#contour detection - which are boundaries of object. contours and edges are two different things

gray = cv.cvtColor(img,cv.COLOR_BGR2GRAY)

#convert image to grey scale

cv.imshow('gray', gray)

#blur

blur =cv.GaussianBlur(gray,(5,5),cv.BORDER_DEFAULT)

cv.imshow('Blur',blur)

canny = cv.Canny(blur,125,175)

#grab the edges

# (125,175) -threshold value

cv.imshow('canny',canny)

contours,hierarchies = cv.findContours(canny,cv.RETR_LIST,cv.CHAIN_APPROX_SIMPLE)

#cv.(something as below) - mode to find contour

#cv.RETR_TREE - all hierarchical contour

#cv.RETR_EXTERNAL - external contour only

#cv.RETR_LIST - all contour

#cv.CHAIN_APPROX_NONE - contour approximation ; take all points

#cv.CHAIN_APPROX_SIMPLE - only takes end points

print(f'{len(contours)} contour(s) found')

cv.waitKey(0)

Result

Notice the number of contours reduced when blurred

Finding edges by threshold method

import cv2 as cv

img = cv.imread('Photos/cat1.jpeg')

cv.imshow('Cats',img)

#contour detection - which are boundaries of object. contours and edges are two different things

gray = cv.cvtColor(img,cv.COLOR_BGR2GRAY)

#convert image to grey scale

cv.imshow('gray', gray)

#threshold  - looks at image and binarize the image

ret,thresh = cv.threshold(gray,125,255,cv.THRESH_BINARY)

#gray - take it gray image

#125 - threshold value

# 255 - maximum value

# Below 125 - set to black

# Above 125 -  set white

#cv.THRESH_BINARY - to binarize image

cv.imshow('thresh', thresh)

contours,hierarchies = cv.findContours(thresh,cv.RETR_LIST,cv.CHAIN_APPROX_SIMPLE)

#cv.(something as below) - mode to find contour

#cv.RETR_TREE - all hierarchical contour

#cv.RETR_EXTERNAL - external contour only

#cv.RETR_LIST - all contour

#cv.CHAIN_APPROX_NONE - contour approximation ; take all points

#cv.CHAIN_APPROX_SIMPLE - only takes end points

print(f'{len(contours)} contour(s) found')

cv.waitKey(0)

Result

Drawing contour on a blank page

- can use blur or threshold method 

- use canny method first and then find the contour using that

- threshold method has disadvantage

import cv2 as cv

import numpy as np

img = cv.imread('Photos/cat1.jpeg')

cv.imshow('Cats',img)

blank = np.zeros(img.shape, dtype ='uint8')

cv.imshow ('Blank', blank )

#contour detection - which are boundaries of object. contours and edges are two different things

gray = cv.cvtColor(img,cv.COLOR_BGR2GRAY)

#convert image to grey scale

cv.imshow('gray', gray)

#threshold  - looks at image and binarize the image

ret,thresh = cv.threshold(gray,125,255,cv.THRESH_BINARY)

#gray - take it gray image

#125 - threshold value

# 255 - maximum value

# Below 125 - set to black

# Above 125 -  set white

#cv.THRESH_BINARY - to binarize image

cv.imshow('thresh', thresh)

contours,hierarchies = cv.findContours(thresh,cv.RETR_LIST,cv.CHAIN_APPROX_SIMPLE)

#cv.(something as below) - mode to find contour

#cv.RETR_TREE - all hierarchical contour

#cv.RETR_EXTERNAL - external contour only

#cv.RETR_LIST - all contour

#cv.CHAIN_APPROX_NONE - contour approximation ; take all points

#cv.CHAIN_APPROX_SIMPLE - only takes end points

print(f'{len(contours)} contour(s) found')

#drawing contour on blank image

cv.drawContours(blank,contours,-1, (0,0,255), 2)

# -1 - how many contours do we want, since we want all of them, its -1

# (0,0,255) - red

# 2 - thickness

cv.imshow('contours', blank)

cv.waitKey(0)

Result

Switch color spaces in OpenCV

1. convert BGR image to HSV

import cv2 as cv

img = cv.imread('Photos/car1.jpeg')

cv.imshow('cat1',img)

# convert BGR to HSV

# HSV - based on how huma think and conceive color

hsv = cv.cvtColor(img, cv.COLOR_BGR2HSV)

cv.imshow('hsv',hsv)

cv.waitKey(0)

Result

Code 2 : Convert BGR to Lab

import cv2 as cv

img = cv.imread('Photos/cat1.jpeg')

cv.imshow('cat1',img)

#convert BGR to Lab

lab = cv.cvtColor(img, cv.COLOR_BGR2Lab)

cv.imshow('lab', lab)

cv.waitKey(0)

Result

Code 3 : convert BGR to RGB

OpenCV reads image in BGR format

In real world(ourside of OpenCV), it uses the RGB format

import cv2 as cv

import matplotlib.pyplot as plt

img = cv.imread('photos/cat1.jpg')

cv.imshow('cat1', img)

#convert BGR to RGB

plt.imshow(img)

plt.show()

Result- RGB in open CV 

Code : Convert BGR to RGB in OpenCV

import cv2 as cv

import matplotlib.pyplot as plt

img = cv.imread('photos/cat1.jpg')

cv.imshow('cat1', img)

#convert BGR to Lab

lab = cv.cvtColor(img, cv.COLOR_BGR2Lab)

cv.imshow('lab', lab)

#BGR to RGB

rgb =cv.cvtColor(img,cv.COLOR_BGR2RGB)

cv.imshow('rgb',rgb)

plt.imshow(rgb)

plt.show()

cv.waitKey(0)

Result : RGB in Open CV

Result : RGB in Matplotlib (outside open cv)

Conversion can be done , but not directly from gray scale to HSV

To convert grayscale to HSV, convert grayscale to BGR, then BGR to HSV

Code : HSV to BGR 

import cv2 as cv

import matplotlib.pyplot as plt

img = cv.imread('photos/cat1.jpg')

cv.imshow('cat1', img)

hsv = cv.cvtColor(img, cv.COLOR_BGR2HSV)

cv.imshow('HSV', hsv)

hsv_bgr = cv.cvtColor(hsv,cv.COLOR_HSV2BGR)

cv.imshow('HSV to BGR', hsv_bgr)

cv.waitKey(0)

Code : LAB to BGR

import cv2 as cv

import matplotlib.pyplot as plt

img = cv.imread('photos/cat1.jpg')

cv.imshow('cat1', img)

#convert BGR to Lab

lab = cv.cvtColor(img, cv.COLOR_BGR2Lab)

cv.imshow('lab', lab)

lab_bgr = cv.cvtColor(lab,cv.COLOR_LAB2BGR)

cv.imshow('lab to BGR', lab_bgr)

cv.waitKey(0)

Color Channels 

- How to split and merge color channels in OpenCV

Code 1 : Split BGR image to color components

import cv2 as cv

import numpy as np

img = cv.imread('photos/cat1.jpg')

cv.imshow('Cat1', img)

b,g,r = cv.split(img)

#cv.split (img) - splits the image to blue, green and red

cv.imshow('Blue',b)

cv.imshow('green', g)

cv.imshow ('red', r )

print (img.shape)

print(b.shape)

print (g.shape)

print(r.shape)

cv.waitKey(0)

Result 

Code 2 : Merge BGR image

import cv2 as cv

import numpy as np

img = cv.imread('photos/cat1.jpg')

cv.imshow('Cat1', img)

b,g,r = cv.split(img)

#cv.split (img) - splits the image to blue, green and red

cv.imshow('Blue',b)

cv.imshow('green', g)

cv.imshow ('red', r )

print (img.shape)

print(b.shape)

print (g.shape)

print(r.shape)

#merge image

merged = cv.merge([b,g,r])

cv.imshow('merged',merged)

cv.waitKey(0)

Code 3 : Extracting image as their colored components

import cv2 as cv

import numpy as np

img = cv.imread('photos/cat1.jpg')

cv.imshow('Cat1', img)

blank = np.zeros(img.shape[:2], dtype ='uint8')

#setting a blank component

b,g,r = cv.split(img)

#cv.split (img) - splits the image to blue, green and red

blue = cv.merge([b, blank,blank])

#setting the green and red components as blank (black)

green = cv.merge ([blank,g,blank])

red = cv.merge ([ blank, blank, r])

cv.imshow('Blue',blue)

cv.imshow('green', green)

cv.imshow ('red', red )

print (img.shape)

print(b.shape)

print (g.shape)

print(r.shape)

#merge image

merged = cv.merge([b,g,r])

cv.imshow('merged',merged)

cv.waitKey(0)

Result

Blurring Techniques

Code 1 : Smoothing image - caused by noise

import cv2 as cv

img = cv.imread('photos/cat1.jpg')

cv.imshow('Cat1', img)

#averaging - get average pixel intensity

average =cv.blur(img,(3,3)) #(3,3)-kernel size which have 3 row 3 column. higher kernel size more blur

cv.imshow('Average Blur', average)

cv.waitKey(0)

Code 2 - Gaussian blur

import cv2 as cv

img = cv.imread('photos/cat1.jpg')

cv.imshow('Cat1', img)

#gaussian blur - less blur

gauss = cv.GaussianBlur (img,(7,7),0)

cv.imshow('Gaussian Blur',gauss)

cv.waitKey(0)

Result 

Code 3 : Median Blur

import cv2 as cv

img = cv.imread('photos/cat1.jpg')

cv.imshow('Cat1', img)

#median blurring - not effetctive for high kernel size. more effective in reducing noise. instead of finding average it finds median

median = cv.medianBlur(img, 3) #assuming size 3X3

cv.imshow('Median Blur',median)

cv.waitKey(0)

Result 

Code 4 : Bilateral blurring

import cv2 as cv

img = cv.imread('photos/cat1.jpg')

cv.imshow('Cat1', img)

#Bilateral blurring - most effective .

bilateral = cv.bilateralFilter(img,5,15 ,15) #5 diameter, 15 colour sigma means more color, 15 space sigma means pixel further from central

cv.imshow('Bilateral',bilateral)

cv.waitKey(0)

Result

BITWISE Operator(and, or , not)
Operate in a binary manner
On - 1
Off - 0

BITWISE AND - takes in common image / intersecting region

blank = np.zeros ((400,400),dtype='uint8')

rectangle = cv.rectangle(blank.copy(),(30,30),(370,370),255,-1) # 30pixel, 255 white,-1 thickness

circle=cv.circle(blank.copy(),(200,200),200,255,-1) #200 centre, 200 radius 255 color,

cv.imshow('Rectangle',rectangle)

cv.imshow('Circle',circle)

#bitwise AND

bitwise_and=cv.bitwise_and(rectangle,circle)#

cv.imshow('BITWISEAND', bitwise_and)

cv.waitKey(0)

BITWISE OR - common region and not common and overlap / non intersecting and intersecting region

bitwise_or = cv.bitwise_or(rectangle,circle)

cv.imshow('Bitwise OR',bitwise_or)

cv.waitKey(0)

BITWISE XOR - non-intecting image 

bitwise_xor =cv.bitwise_xor(rectangle,circle)

cv.imshow('bitwise XOR',bitwise_xor)

Note : XOR -OR = AND

AND-OR =XOR

BITWISE (NOT)

bitwise_not =cv.bitwise_not(rectangle,circle)

cv.imshow('bitwise NOT', bitwise_not)

FACE DETECTION - haarcascade
- detection of face
- uses classifier -> present or not
- OPENCV uses predefined classifier -1. haarcascade 2. local binary pattern- advanced

1. Open haarcascade_frontalface_default_xml 

2. Click Raw > ctrl A > ctrl C

3. Create new file >name it "haar_face.xml" > paste the code > save  

1. FACE DETECTION - Number of faces

import cv2 as cv

img=cv.imread('Photos/tom4.jpeg')

cv.imshow ('tom',img)

#convert to grey scale

gray = cv.cvtColor(img,cv.COLOR_BGR2GRAY)

cv.imshow('GRAY', gray)

haar_cascade =cv.CascadeClassifier ('haar_face.xml')

faces_rect = haar_cascade.detectMultiScale(gray,scaleFactor = 1.1, minNeighbors=3)

print (f'Number of faces = {len(faces_rect)}')

cv.waitKey(0)

RESULT 

2. Draw a rectangle over image 

#to draw a rectangle over the detected image

for(x,y,w,h) in faces_rect: #x,y,w,h >

    cv.rectangle(img,(x,y), (x+w,y+h), (0,255,0), thickness=2) #cv.rectangle draw rectangle, wyhw coordinate

cv.imshow('Detected faces', img)

RESULT 

GROUP PIC example

import cv2 as cv

img=cv.imread('Photos/g3.jpg')

cv.imshow ('g3',img)

#convert to grey scale

gray = cv.cvtColor(img,cv.COLOR_BGR2GRAY)

cv.imshow('GRAY', gray)

haar_cascade =cv.CascadeClassifier ('haar_face.xml')

faces_rect = haar_cascade.detectMultiScale(gray,scaleFactor = 2.2, minNeighbors=3)

print (f'Number of faces = {len(faces_rect)}')

#to draw rectangle over detected image

for(x,y,w,h) in faces_rect: #x,y,w,h >

    cv.rectangle(img,(x,y), (x+w,y+h), (0,255,0), thickness=2) #cv.rectangel draw rectangle, wyhw coordinate

cv.imshow('Detected faces', img)

cv.waitKey(0)

RESULT

To increase accuracy increase the minimum neighbours
Pros/cons of haar cascade
1. Popular/not advanced
2. Les set up /Dealeds face recognizer
3. can be used for video

FACE RECOGNITION - with  OpenCV's built-in recognizer - train the recognizer

1. Safe at least 15 images of 3 types in a folder

import os

import cv2 as cv

import numpy as np

people =[ 'George Harrison, Ozil ']

RESULT