model_prediction.py

import cv2
import os
from get_data import *
from model_generation import *

#Image captured from a live video source. Reshaped and used to predict
#A timer array keeps count of the emotions and returns the most shown emotion after 10 seconds
def resource_path(relative_path):
    try:
        base_path = sys._MEIPASS
    except Exception:
        base_path = os.path.abspath(".")

    return os.path.join(base_path, relative_path)

def timer(emotion, timer_array):
    count = timer_array.get(emotion)
    count += 1
    timer_array[emotion] = count
    return timer_array

def reshape_image(image):
    requiredsize = (48, 48)
    image = cv2.resize(image, requiredsize, interpolation=cv2.INTER_AREA)
    image = image.reshape(-1, image.shape[0], image.shape[1], 1)
    image = image.astype('float32')/32
    return image

def predict(image, model, timer_array):
    output = model.predict(image)
    index = np.argmax(output)
    emotion = ["Angry", "Disgust", "Fear", "Happy", "Sad", "Suprise", "Neutral"]
    color = [(0, 0, 255), (255, 0, 255), (0, 0, 255), (0, 255, 0), (0, 255, 255), (0, 255, 255), (0, 255, 0)]
    timer_array = timer(emotion[index], timer_array)
    return emotion[index], color[index], timer_array

def get_face(model):
    os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3' 
    # cv2_base_dir = os.path.dirname(resource_path(cv2.__file__))
    # haar_model = os.path.join(cv2.data.haarcascades + 'haarcascade_frontalface_default.xml')
    cv2.ocl.setUseOpenCL(False)

    timer_array = {"Angry":0, "Disgust":0, "Fear":0, "Happy":0, "Sad":0, "Suprise":0, "Neutral":0}

    live_video = cv2.VideoCapture(0)
    
    while True:
        _, frame = live_video.read()
        frame = cv2.flip(frame, flipCode=1)

        face_cascade = cv2.CascadeClassifier(resource_path("Extras/haarcascade_frontalface_default.xml"))
        faces = face_cascade.detectMultiScale(frame, scaleFactor=1.2, minNeighbors=5, minSize=(20, 20))

        for (x, y, w, h) in faces:
            cv2.rectangle(frame, (x, y), (x+w, y+h), (0, 255, 0), 0)

            face = frame[y:y+h, x:x+w]
            face = cv2.cvtColor(face, cv2.COLOR_BGR2GRAY)
            face = reshape_image(face)

            text, color, timer_array = predict(face, model, timer_array)
            
            cv2.rectangle(frame, (x, y), (x+w, y+h), color, 2)
            cv2.rectangle(frame, (x+w-150, y+h), (x+w, y+h+30), color, -1)
            cv2.putText(frame, text, (x+w-100, y+h+20), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (255, 255, 255), 2)
        cv2.imshow("Feed", frame)

        if cv2.waitKey(1) & 0xFF == ord('q'): 
            break

    live_video.release()
    cv2.destroyAllWindows()

    average_emotion = max(timer_array, key=timer_array.get) 
    return average_emotion