image_recognition.py

import os
import numpy as np
from PIL import Image,ImageFont,ImageDraw
import matplotlib.pyplot as plt
import cv2
from script.fx import prewhiten,l2_normalize
from keras.models import load_model
from scipy.spatial import distance

model_path='./data/model/facenet_keras.h5'
face_cascade_path='./data/cascade/haarcascade_frontalface_default.xml'
font_path='../data/font/Calibri Regular.ttf'
embedding_path='./data/arrays/embeddings.npz'
vars_path='./data/arrays/vars.npz'

model=load_model(model_path)
face_cascade=cv2.CascadeClassifier(face_cascade_path)
loaded_embeddings=np.load(embedding_path)
embeddings,names=loaded_embeddings['a'],loaded_embeddings['b']
loaded_vars=np.load(vars_path)
slope,intercept=loaded_vars['a'],loaded_vars['b']


os.system('cls')
print("\n\n\nKeep the images in './test/' directory.")
input("Press ENTER when you're ready.")


os.chdir('./test/')
dr=os.listdir()
dr.remove('predicted')
for c in dr:
    frame=cv2.imread(c)
    gray=cv2.cvtColor(frame,cv2.COLOR_BGR2GRAY)
    faces=face_cascade.detectMultiScale(gray,scaleFactor=1.2,minNeighbors=10)
    
    for (x_face,y_face,w_face,h_face) in faces:
        
        #Margins for Face box
        dw=0.1*w_face
        dh=0.2*h_face
        
        dist=[]
        for i in range(len(embeddings)):
            dist.append(distance.euclidean(l2_normalize(model.predict(prewhiten(cv2.resize(frame[y_face:y_face+h_face,x_face:x_face+w_face],(160,160)).reshape(-1,160,160,3)))),embeddings[i].reshape(1,128)))
        dist=np.array(dist)
        if dist.min()>1:
            name='Unidentified'
        else:   
            name=names[dist.argmin()]
        print(name,dist.min())
        
        if name!='Unidentified':
            font_size=int(slope[dist.argmin()]*((w_face+2*dw)//3)*2+intercept[dist.argmin()])
        else:
            font_size=int(0.1974311*((w_face+2*dw)//3)*2+0.03397702412218706)
            
        font=ImageFont.truetype(font_path,font_size)
        size=font.getsize(name)

        cv2.rectangle(frame,
                      pt1=(x_face-int(np.floor(dw)),(y_face-int(np.floor(dh)))),
                      pt2=((x_face+w_face+int(np.ceil(dw))),(y_face+h_face+int(np.ceil(dh)))),
                      color=(0,255,0),
                      thickness=2) #Face Rectangle
        
        cv2.rectangle(frame,
                      pt1=(x_face-int(np.floor(dw)),y_face-int(np.floor(dh))-size[1]),
                      pt2=(x_face-int(np.floor(dw))+size[0],y_face-int(np.floor(dh))),
                      color=(0,255,0),
                      thickness=-1) #Face Name background rectangle
        
        img=Image.fromarray(frame)
        draw=ImageDraw.Draw(img)
        draw.text((x_face-int(np.floor(dw)),y_face-int(np.floor(dh))-size[1]),name,font=font,fill=(255,0,0))
        frame=np.array(img)

    frame=cv2.cvtColor(frame,cv2.COLOR_BGR2RGB)
    plt.imsave('./predicted/'+c,frame)
print("\n\n\nOutputs can be found in './test/predicted/'")