detectfaces.py

# Function to detect faces in a grascale image returns its top right corner and resized image of 48x48
# Author: Shashi Kant
# Date: 07/12/2018

import numpy as np
import cv2

# for haar cascade
face_cascade = cv2.CascadeClassifier('saved_model/haarcascade_frontalface_default.xml')

# for dnn based face detection
modelFile = "saved_model/opencv_face_detector_uint8.pb"
configFile = "saved_model/opencv_face_detector.pbtxt"

#net = cv2.dnn.readNetFromTensorflow(modelFile, configFile)

'''def dnn_faces(img, thres):
    blob = cv2.dnn.blobFromImage(cv2.resize(img, (300,300)), 1.0, (300, 300), [104, 117, 123])
    net.setInput(blob)
    predict = net.forward()

    faces_detected = []
    [h, w] = img.shape[:2]
    for i in range(predict.shape[2]):
        confidence = predict[0, 0, i, 2]
        if confidence > thres:
            x1 = predict[0, 0, i, 3] * w
            y1 = predict[0, 0, i, 4] * h
            x2 = predict[0, 0, i, 5] * w
            y2 = predict[0, 0, i, 6] * h
            wf = x2 - x1
            hf = y2 - y1
            faces_detected.append((int(x1), int(y1), int(wf), int(hf)))

    return faces_detected'''

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