lipReading.py

# coding: utf-8
import sys
import dlib
import cv2
import numpy as np
import os


def file_name(file_dir):
    L = []
    for root, dirs, files in os.walk(file_dir):
        for file in files:
            if os.path.splitext(file)[1] == '.mpg':
                L.append(os.path.join(root, file))
    return L


detector = dlib.get_frontal_face_detector()  # 获取人脸分类器
predictor = dlib.shape_predictor('shape_predictor_68_face_landmarks.dat')
cnt = 0  # count videos
while (1):
    videos_path = file_name("s1")
    if cnt >= len(videos_path):
        print("no more videos!")
        break
    cap = cv2.VideoCapture(videos_path[cnt])
    cnt = cnt + 1
    pics_num = 0  # count frames
    ########################################################To write lip images as video#######################################

    # fps = cap.get(cv2.CAP_PROP_FPS)
    # size = (int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)),int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)))
    # video_saver = cv2.VideoWriter("test",cv2.cv2.CAP_PROP_FOURCC('M', 'J', 'P', 'G'),fps,size)
    # still have some problem on this
    ############################################################################################################################
    while (1):
        succ, img = cap.read()
        if img is None:
            break
        rects = detector(img, 0)
        # pics_num = 0
        for i in range(len(rects)):
            landmarks = np.matrix([[p.x, p.y] for p in predictor(img, rects[i]).parts()])
            top = 100000
            bottle = 0
            left = 10000
            right = 0
            for idx, point in enumerate(landmarks):
                if idx > 48:
                    pos = (point[0, 0], point[0, 1])
                    if point[0, 0] < left:
                        left = point[0, 0]
                    if point[0, 0] > right:
                        right = point[0, 0]
                    if point[0, 1] < top:
                        top = point[0, 1]
                    if point[0, 1] > bottle:
                        bottle = point[0, 1]
                    cv2.circle(img, pos, 1, color=(0, 255, 0))
            w_offset = (right - left) // 4
            h_offset = (bottle - top) // 2
            cv2.rectangle(img, (left - w_offset, top - h_offset), (right + w_offset, bottle + h_offset), (0, 0, 255), 3)
            liproi = img[top - h_offset:bottle + h_offset, left - w_offset:right + w_offset]
            # print ((left-w_offset,top-h_offset),(right+w_offset,bottle+h_offset))

        # video_saver.write(img)           #write video
        cv2.imwrite("pics/" + str(cnt) + "/" + str(pics_num) + ".jpg", liproi)
        print(cnt, pics_num)
        pics_num = pics_num + 1
        cv2.imshow("", img)
        cv2.imshow("lip", liproi)
        cv2.waitKey(1)