face_recog.py

import cv2
import sys
import os
import re
import numpy as np
import shelve,random

FONT = cv2.FONT_HERSHEY_SIMPLEX
CASCADE = "face_cascade.xml"
FACE_CASCADE = cv2.CascadeClassifier(CASCADE)

Datafile = shelve.open("Data")
if 'Data' not in Datafile.keys():
    Datafile['Data']=list()
    Data_list = list()
else:
    Data_list = Datafile["Data"]


def Make_Changes(label):
    if label not in Data_list:
        Data_list.append(label)
	print Data_list


def get_images(path):
    images = list()
    labels = list()
    count=0
    if len(os.listdir(path)) == 0:
        print "Empty Dataset.......aborting Training"
        exit()
    for img in os.listdir(path):
        regex = re.compile(r'(\d+|\s+)')
        labl = regex.split(img)
        labl = labl[0]
        count=count+1
        Make_Changes(labl)
        image_path =os.path.join(path,img)
        image=cv2.imread(image_path)
        image_grey=cv2.cvtColor(image,cv2.COLOR_BGR2GRAY)
        images.append(image_grey)
        labels.append(Data_list.index(labl))
    return images,labels,count


#def add_to_dataset(image):
def initialize_recognizer():
    try:
        face_recognizer = cv2.face.createLBPHFaceRecognizer()
    except:
        face_recognizer = cv2.createLBPHFaceRecognizer()
    print "Training.........."
    Dataset = get_images("./Dataset")
    print "Recognizer trained using Dataset: "+str(Dataset[2])+" Images used"
    face_recognizer.train(Dataset[0],np.array(Dataset[1]))
    return face_recognizer


def save_wrong_faces(num,temp_set,faces):
    os.chdir("./Dataset")
    if num:
        print "Enter number below face : Correct Name:"
        for i in xrange(num):
            inp = raw_input()
            inp = inp.split(":")
            faces[int(inp[0])][0]=-1
            if(inp[1].lower() != "nil"):
                cv2.imwrite(inp[1]+ str(random.uniform(0,100000))+ ".jpg",temp_set[int(inp[0])])
    for i in xrange(len(faces)):
        if faces[i][0]!=-1 and faces[i][1]>18:
            cv2.imwrite(Data_list[faces[i][0]]+str(random.uniform(0,100000))+ ".jpg",temp_set[i])
    os.chdir("../")


def recognize(image_path,face_recognizer):
    image=cv2.imread(image_path)
    try:
        image_grey=cv2.cvtColor(image,cv2.COLOR_BGR2GRAY)
    except:
        print "Image not Found"
        exit()
    faces = FACE_CASCADE.detectMultiScale(image_grey,scaleFactor=1.16,minNeighbors=5,minSize=(25,25),flags=0)
    temp_set = list()
    face_list = list()
    num=0
    for x,y,w,h in faces:
        sub_img=image_grey[y:y+h,x:x+w]
        img=image[y:y+h,x:x+w]
        temp_set.append(img)
        nbr,conf = face_recognizer.predict(sub_img)
        face_list.append([nbr,conf]);
        cv2.rectangle(image,(x-5,y-5),(x+w+5,y+h+5),(255, 255,0),2)
        cv2.putText(image,Data_list[nbr],(x,y-10), FONT, 0.5,(255,255,0),1)
        cv2.putText(image,str(num),(x,y+h+20), FONT, 0.5,(255,255,0),1)
        cv2.imwrite("Detected.jpg",image)
        num = int(num)+1
    Datafile["Data"]=Data_list
    Datafile.close()
    os.system("xdg-open Detected.jpg")
    print "No. of faces predicted wrong:"
    num_wrong = input()
    save_wrong_faces(num_wrong,temp_set,face_list)


def recognize_video(face_recognizer):
	cap = cv2.VideoCapture(0)
	while True:
		if cap.grab():
			ref,image = cap.retrieve()
			image_grey=cv2.cvtColor(image,cv2.COLOR_BGR2GRAY)
			faces = FACE_CASCADE.detectMultiScale(image_grey,scaleFactor=1.16,minNeighbors=5,minSize=(25,25),flags=0)
			for x,y,w,h in faces:
        			sub_img=image_grey[y:y+h,x:x+w]
        			img=image[y:y+h,x:x+w]
        			nbr,conf = face_recognizer.predict(sub_img)
        			cv2.rectangle(image,(x-5,y-5),(x+w+5,y+h+5),(255, 255,0),2)
        			cv2.putText(image,Data_list[nbr],(x,y-10), FONT, 0.5,(255,255,0),1)			
        		cv2.imshow("Faces Found",image)
		if (cv2.waitKey(1) & 0xFF == ord('q')) or (cv2.waitKey(1) & 0xFF == ord('Q')):
			break
	Datafile["Data"]=Data_list
    	Datafile.close()
	cap.release()
	cv2.destroyAllWindows()


def main():
    if len(sys.argv) not in [1, 2]:
        print "Usage: python face_recog.py [<complete image_path>]"
        sys.exit()

    face_r = initialize_recognizer()

    if len(sys.argv) == 1:
        recognize_video(face_r)
    else:
        recognize(sys.argv[1], face_r)


if __name__ == "__main__":
    main()