IOSD_Class4_face_recog.py

import numpy as np
import cv2

cam=cv2.VideoCapture(0)
facec=cv2.CascadeClassifier('haarcascade_frontalface_default.xml')
font = cv2.FONT_HERSHEY_SIMPLEX

# Load data in train (X and y)

f_01 = np.load('face01.npy').reshape((20, -1))
f_02 = np.load('face04.npy').reshape((20, -1))
f_03 = np.load('face03.npy').reshape((20, -1))

data = np.concatenate((f_01, f_02,f_03))
labels =  np.zeros((data.shape[0],))

labels[20:40] = 1.0
labels[40:] = 2.0

names = {
	0: 'Rajat',
	1: 'Rishabh',
	2: 'Jatin'
}

# Define KNN functions

def distance(x1, x2):
	d = np.sqrt(((x1-x2)**2).sum())
	return d

def knn(X_train, y_train, xt, k=5):
	vals = []
	for ix in range(X_train.shape[0]):
		d = distance(X_train[ix], xt)
		vals.append([d, y_train[ix]])
	sorted_labels = sorted(vals, key=lambda z: z[0])
	neighbours = np.asarray(sorted_labels)[:k, -1]
	
	freq = np.unique(neighbours, return_counts=True)
	
	return freq[0][freq[1].argmax()]

# Run the main loop





while True:
    ret,fr=cam.read()
    if ret == True:
        gray= cv2.cvtColor(fr,cv2.COLOR_BGR2GRAY)  #bgr bcz it bydefault loads image in bgr format instead of rgb
        faces=facec.detectMultiScale(gray,1.3,5)


        
        for (x,y,w,h) in faces:
            # Extract detected face

                fc=fr[y:y+h,x:x+w,:]  #bcz here rows are on the y axis that's why we use y first then x for image processing
                # resize to a fixed shape
                r = cv2.resize(fc,(50,50)).flatten()  #here x is the no of colums then rows
                text = names[int(knn(data, labels, r))]
                cv2.putText(fr, text, (x, y), font, 1, (255, 255, 0), 2)
                cv2.rectangle(fr,(x,y),(x+w,y+w),(0,0,255),2)  #image values are of int always bcz of pixel move from 0 to 255
        

        cv2.imshow('frame',fr)
        if cv2.waitKey(1)==27 :
            break
    else:
        print ("error")
        break
    
cv2.destroyAllWindows()