haar_cascade.py

# Haar cascade face extraction
import os
from cv2 import cv2, data
from glob import glob
from util import image_extensions, has_img_extension

imgdir = os.getcwd()
save_dir = 'extracted'
ext = image_extensions  # Image formats


def find_images_in_dir(imgdir):
    all_files = os.listdir(imgdir)
    files = [file for file in all_files if has_img_extension(file)]
    return files


files = find_images_in_dir(imgdir)
filenames = [os.path.splitext(filepath) for filepath in files]

images = [cv2.imread(file) for file in files]
figure_size = (160, 160)


def haar_cascade_extraction(images):

    for index, image in enumerate(images):
        grayscaled_img = cv2.cvtColor(
            image, cv2.COLOR_BGR2GRAY)  # Convert to grayscale
        face_cascade = cv2.CascadeClassifier(
            data.haarcascades + "haarcascade_frontalface_alt2.xml")
        faces = face_cascade.detectMultiScale(grayscaled_img, 1.1, 4)

        for count, (x, y, w, h) in enumerate(faces):
            # Draw rectangle around face
            cv2.rectangle(image, (x, y), (x+w, y+h), (255, 255, 255))
            face = image[y:y + h, x:x + w]
            resized = cv2.resize(face, figure_size)
            # Counts up if there is more than one face in image

            print(f'Cropping and Saving: {filename}')
            cv2.imshow('', resized)
            cv2.imwrite(filename, resized)
            cv2.waitKey(1)
            cv2.destroyAllWindows()


if __name__ == '__main__':
    folder = 'men./'
    current_path = os.getcwd()
    if not current_path.endswith(folder):
        os.chdir(folder)
        haar_cascade_extraction(images)