Automatic coding - but in python

some nonsense written on the front

When we watch videos, sometimes there are some strange mosaics that affect our viewing experience, so how exactly are these mosaics added?

This time we will use Python to automatically encode videos!

Ready to work

We can still use Python3.8 and pycharm2021 for the environment

Implementation principle

1. Divide video into audio and picture;
2. The face appears in the screen and the target is compared, and the corresponding face is coded;
3. Add sound to the processed video;

module

• Manually install the cv2 module, pip install opencv-python installation
• The installation encounters an error, and it will not be installed. See the article on the top of my homepage.

material tool

We need to install the ffmpeg audio and video transcoding tool

All materials can be obtained by scanning the code on the left

Code parsing

Import required modules

import cv2  
import face_recognition  # 人脸识别库  99.7%    cmake  dlib  face_recognition
import subprocess

Convert video to audio

def video2mp3(file_name):
    """
    :param file_name: 视频文件路径
    :return:
    """
    outfile_name = file_name.split('.')[0] + '.mp3'
    cmd = 'ffmpeg -i ' + file_name + ' -f mp3 ' + outfile_name
    print(cmd)
    subprocess.call(cmd, shell=False)

add mosaic

def mask_video(input_video, output_video, mask_path='mask.jpg'):
    """
    :param input_video: 需打码的视频
    :param output_video: 打码后的视频
    :param mask_path: 打码图片
    :return:
    """
    # 读取图片
    mask = cv2.imread(mask_path)
    # 读取视频
    cap = cv2.VideoCapture(input_video)
    # 视频  fps  width  height
    v_fps = cap.get(5)
    v_width = cap.get(3)
    v_height = cap.get(4)

    # 设置写入视频参数  格式MP4
    # 画面大小
    size = (int(v_width), int(v_height))
    fourcc = cv2.VideoWriter_fourcc('m', 'p', '4', 'v')

    # 输出视频
    out = cv2.VideoWriter(output_video, fourcc, v_fps, size)

    # 已知人脸
    known_image = face_recognition.load_image_file('tmr.jpg')
    biden_encoding = face_recognition.face_encodings(known_image)[0]

    cap = cv2.VideoCapture(input_video)

    while (cap.isOpened()):
        ret, frame = cap.read()
        if ret:
            # 检测人脸
            # 人脸区域
            face_locations = face_recognition.face_locations(frame)

            for (top_right_y, top_right_x, left_bottom_y, left_bottom_x) in face_locations:
                print((top_right_y, top_right_x, left_bottom_y, left_bottom_x))
                unknown_image = frame[top_right_y - 50:left_bottom_y + 50, left_bottom_x - 50:top_right_x + 50]
                if face_recognition.face_encodings(unknown_image) != []:
                    unknown_encoding = face_recognition.face_encodings(unknown_image)[0]

                    # 对比人脸
                    results = face_recognition.compare_faces([biden_encoding], unknown_encoding)
                    # [True]
                    # 贴图
                    if results == [True]:
                        mask = cv2.resize(mask, (top_right_x - left_bottom_x, left_bottom_y - top_right_y))
                        frame[top_right_y:left_bottom_y, left_bottom_x:top_right_x] = mask
            out.write(frame)


        else:
            break

Add audio to the picture

def video_add_mp3(file_name, mp3_file):
    """
    :param file_name: 视频画面文件
    :param mp3_file:  视频音频文件
    :return:
    """
    outfile_name = file_name.split('.')[0] + '-f.mp4'
    subprocess.call('ffmpeg -i ' + file_name + ' -i ' + mp3_file + ' -strict -2 -f mp4 ' + outfile_name, shell=False)

full code

import cv2 
import face_recognition  # 人脸识别库  99.7%    cmake  dlib  face_recognition
import subprocess

def video2mp3(file_name):

    outfile_name = file_name.split('.')[0] + '.mp3'
    cmd = 'ffmpeg -i ' + file_name + ' -f mp3 ' + outfile_name
    print(cmd)
    subprocess.call(cmd, shell=False)


def mask_video(input_video, output_video, mask_path='mask.jpg'):

    # 读取图片
    mask = cv2.imread(mask_path)
    # 读取视频
    cap = cv2.VideoCapture(input_video)
    # 视频  fps  width  height
    v_fps = cap.get(5)
    v_width = cap.get(3)
    v_height = cap.get(4)

    # 设置写入视频参数  格式MP4
    # 画面大小
    size = (int(v_width), int(v_height))
    fourcc = cv2.VideoWriter_fourcc('m', 'p', '4', 'v')

    # 输出视频
    out = cv2.VideoWriter(output_video, fourcc, v_fps, size)

    # 已知人脸
    known_image = face_recognition.load_image_file('tmr.jpg')
    biden_encoding = face_recognition.face_encodings(known_image)[0]

    cap = cv2.VideoCapture(input_video)

    while (cap.isOpened()):
        ret, frame = cap.read()
        if ret:
            # 检测人脸
            # 人脸区域
            face_locations = face_recognition.face_locations(frame)

            for (top_right_y, top_right_x, left_bottom_y, left_bottom_x) in face_locations:
                print((top_right_y, top_right_x, left_bottom_y, left_bottom_x))
                unknown_image = frame[top_right_y - 50:left_bottom_y + 50, left_bottom_x - 50:top_right_x + 50]
                if face_recognition.face_encodings(unknown_image) != []:
                    unknown_encoding = face_recognition.face_encodings(unknown_image)[0]

                    # 对比人脸
                    results = face_recognition.compare_faces([biden_encoding], unknown_encoding)
                    # [True]
                    # 贴图
                    if results == [True]:
                        mask = cv2.resize(mask, (top_right_x - left_bottom_x, left_bottom_y - top_right_y))
                        frame[top_right_y:left_bottom_y, left_bottom_x:top_right_x] = mask
            out.write(frame)


        else:
            break


def video_add_mp3(file_name, mp3_file):

    outfile_name = file_name.split('.')[0] + '-f.mp4'
    subprocess.call('ffmpeg -i ' + file_name + ' -i ' + mp3_file + ' -strict -2 -f mp4 ' + outfile_name, shell=False)


if __name__ == '__main__':
    # 1.
    video2mp3('cut.mp4')
    # 2.
    mask_video(input_video='cut.mp4',output_video='output.mp4')
    # 3.
    video_add_mp3(file_name='output.mp4',mp3_file='cut.mp3')

no more~