emotion_recogniser.py

import numpy as np
import cv2
#from detectfaces import get_faces
from keras.models import load_model
from keras.models import Sequential
from keras.layers import Dense, Dropout, Flatten, BatchNormalization
from keras.layers import Conv2D, MaxPooling2D, Activation
from keras import backend as K
import matplotlib.pyplot as plt

img_rows, img_cols = 48, 48
face_cascade = cv2.CascadeClassifier('saved_model/haarcascade_frontalface_default.xml')
modelFile = "saved_model/opencv_face_detector_uint8.pb"
configFile = "saved_model/opencv_face_detector.pbtxt"



# 0=Angry, 1=Disgust, 2=Fear, 3=Happy, 4=Sad, 5=Surprise, 6=Neutral
emotion = ['Angry', "Disgust", 'Fear', 'Happy', 'Sad', 'Surprise', 'Neutral']
font = cv2.FONT_HERSHEY_SIMPLEX

model = []
print('Loading Models...')
print('0/3')
for i in range(2):
    m = load_model('saved_model/' + 'cnn'+str(i)+'.h5')
    print(str(i+1) + '/3')
    model.append(m)

m = load_model('saved_model/ensemble.h5')
model.append(m)
print('3/3')

print("Loading Complete!")

def predict(x):
    x_rev = np.flip(x, 1)
    x = x.astype('float32')
    x_rev = x_rev.astype('float32')
    x /= 255
    x_rev /= 255
    p = np.zeros((1, 14))
    p[:,0:7] = model[0].predict(x.reshape(1,48,48,1))
    p[:,7:14] = model[1].predict(x_rev.reshape(1,48,48,1))
    pre = model[2].predict(p)
    return pre

cap = cv2.VideoCapture(0)
if not cap.isOpened():
    cap.open()

def get_faces(img, method='haar'):
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    faces = []

    #if (method == 'dnn'):
        #faces_detected = dnn_faces(img, 0.5)
    if (method == 'haar'):
        faces_detected = face_cascade.detectMultiScale(gray, 1.3, 5)

    for i, (x,y,w,h) in enumerate(faces_detected):
        my = int(y + h/2)
        mx = int(x + w/2)

        if h<w:
            c = int(h/2)
        else:
            c = int(w/2)

        face = gray[my-c:my+c, mx-c:mx+c]
        try:
            face_48 = cv2.resize(face,(48, 48), interpolation = cv2.INTER_CUBIC)
            faces.append((y, x + w, face_48))
        except:
            pass

    return faces

while(True):
    ret, img = cap.read()

    #faces = get_faces(img, method='dnn')
    faces = get_faces(img, method='haar')

    for i, (y,x,face) in enumerate(faces):
        pre = predict(face)
        k = np.argmax(pre)
        txt = emotion[k] + ' [' + str(int(pre[0,k]*100)) + '%]'
        cv2.putText(img, txt, (x, y), font, 1.0,(0,255,0),2,cv2.LINE_AA)
        # cv2.imshow(str(i), face)

    cv2.imshow('Camera', img)
    if cv2.waitKey(20) & 0xFF == ord('q'):
        break

cap.release()
cv2.destroyAllWindows()