Docker基础知识 (17) - 使用 Docker 部署 Python + OpenCV-Python/Moviepy 编辑图像和视频
Python 是一种由 Guido van Rossum 开发的通用编程语言，它很快就变得非常流行，主要是因为它的简单性和代码可读性。它使程序员能够用更少的代码行表达思想，而不会降低可读性。
Python 可以轻松使用 C/C++ 扩展，这使我们可以在 C/C++ 中编写计算密集型代码，并创建可用作 Python 模块的 Python 包装器。这给我们带来了两个好处：首先，代码与原始 C/C++ 代码一样快（因为它在后台工作的实际 c/C++ 代码），其次，在 Python 中编写代码比使用 C/C++ 更容易。
Python: https://www.python.org/
OpenCV
OpenCV（Open Source Computer Vision）由英特尔公司于 1999 年推出，如今由 Willow Garage 提供支持。它是一个基于 BSD 许可（开源）发行的跨平台计算机视觉库，可以运行在 Linux、Windows、MacOS 操作系统上。
OpenCV 轻量级而且高效，由一系列 C 函数和少量 C++ 类构成，同时提供了 Python、Ruby、MATLAB 等语言的接口，实现了图像处理和计算机视觉方面的很多通用算法，它在 Python 中也被广泛用于计算机视觉。
简言之，通过 OpenCV 可实现计算机图像、视频的编辑。广泛应用于图像识别、运动跟踪、机器视觉等领域。
OpenCV: https://opencv.org/
OpenCV GitHub: https://github.com/opencv
OpenCV-Python
OpenCV-Python 是一个 Python 库/包，旨在解决计算机视觉问题。OpenCV-Python 是原始 OpenCV C/C++ 实现的 Python 库/包。
OpenCV-Python 使用 Numpy，这是一个高度优化的数据库操作库，具有 MATLAB 风格的语法。所有 OpenCV 数组结构都转换为 Numpy 数组。这也使得与使用 Numpy 的其他库（如 SciPy 和 Matplotlib ）集成更容易。
OpenCV-Python: https://pypi.org/project/opencv-python
MoviePy
MoviePy 是一个用于视频编辑的 Python 模块，可用于进行视频的基本操作（如剪切、连接、标题插入）、视频合成（也称非线性编辑）、视频处理或创建高级效果。
MoviePy 能处理的视频是 ffmpeg 格式的，支持的文件类型：*.mp4 *.wmv *.rm *.avi *.flv *.webm *.wav *rmvb 。
MoviePy: https://pypi.org/project/moviepy/
Github: https://github.com/Zulko/moviepy
注：Moviepy 是一个 Python 的音视频剪辑库，OpenCV 是一个图形处理库，视频的一帧就是一幅图像，因此在处理视频时可以结合 OpenCV 进行帧处理，将二者结合可以用来进行视频特效的处理。
1. 部署环境
IP 地址（本地测试环境）：192.168.0.10
操作系统：Linux CentOS 7.9
Docker 版本： 20.10.7
Docker Compose 版本： 2.6.1
OpenCV-Python 版本：3.4.8.29
工作目录：/home/docker/python_media
2. 创建 Dockerfile
$ cd /home/docker/python_media
$ vim Dockerfile
FROM python:3.8
WORKDIR /home/docker/python_media/work
#COPY . /home/docker/python_media/work
RUN pip install -i https://mirrors.aliyun.com/pypi/simple opencv-python==3.4.8.29
RUN pip install -i https://mirrors.aliyun.com/pypi/simple moviepy==1.0.3
3. 创建 python_media 镜像
$ cd /home/docker/python_media
$ docker build -t python_media:3.8 .
Sending build context to Docker daemon 9.216kB
Step 1/4 : FROM python:3.8
---> 5e51aed29a27
Step 2/4 : WORKDIR /home/docker/python_media
---> Running in bfa5952eec12
Removing intermediate container bfa5952eec12
---> 536afc6d0b85
Step 3/4 : RUN pip install -i https://mirrors.aliyun.com/pypi/simple opencv-python==3.4.8.29
---> Running in 4238ef77d8e2
Looking in indexes: https://mirrors.aliyun.com/pypi/simple
Collecting opencv-python==3.4.8.29
Downloading https://mirrors.aliyun.com/pypi/packages/bd/95/f06beabb5e1fa319f02278004efce19e7ebe7a67e69dd7d1820d64b2dab8/opencv_python-3.4.8.29-cp38-cp38-manylinux1_x86_64.whl (28.3 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 28.3/28.3 MB 261.3 kB/s eta 0:00:00
Collecting numpy>=1.17.3
Downloading https://mirrors.aliyun.com/pypi/packages/56/df/2f6016171ebce9875e7de0292a2131bea86e0340607a313a04b332d35c8e/numpy-1.23.4-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (17.1 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 17.1/17.1 MB 257.3 kB/s eta 0:00:00
Installing collected packages: numpy, opencv-python
...
Step 4/4 : RUN pip install -i https://mirrors.aliyun.com/pypi/simple moviepy==1.0.3
---> Running in 4011b24e7bc4
Looking in indexes: https://mirrors.aliyun.com/pypi/simple
Collecting moviepy==1.0.3
Downloading https://mirrors.aliyun.com/pypi/packages/18/54/01a8c4e35c75ca9724d19a7e4de9dc23f0ceb8769102c7de056113af61c3/moviepy-1.0.3.tar.gz (388 kB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 388.3/388.3 KB 201.0 kB/s eta 0:00:00
Preparing metadata (setup.py): started
Preparing metadata (setup.py): finished with status 'done'
Collecting decorator<5.0,>=4.0.2
Downloading https://mirrors.aliyun.com/pypi/packages/ed/1b/72a1821152d07cf1d8b6fce298aeb06a7eb90f4d6d41acec9861e7cc6df0/decorator-4.4.2-py2.py3-none-any.whl (9.2 kB)
Collecting tqdm<5.0,>=4.11.2
Downloading https://mirrors.aliyun.com/pypi/packages/47/bb/849011636c4da2e44f1253cd927cfb20ada4374d8b3a4e425416e84900cc/tqdm-4.64.1-py2.py3-none-any.whl (78 kB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 78.5/78.5 KB 200.4 kB/s eta 0:00:00
Collecting requests<3.0,>=2.8.1
Downloading https://mirrors.aliyun.com/pypi/packages/ca/91/6d9b8ccacd0412c08820f72cebaa4f0c0441b5cda699c90f618b6f8a1b42/requests-2.28.1-py3-none-any.whl (62 kB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 62.8/62.8 KB 189.6 kB/s eta 0:00:00
Collecting proglog<=1.0.0
Downloading https://mirrors.aliyun.com/pypi/packages/8b/f5/cab5cf6a540c31f5099043de0ae43990fd9cf66f75ecb5e9f254a4e4d4ee/proglog-0.1.10-py3-none-any.whl (6.1 kB)
Requirement already satisfied: numpy>=1.17.3 in /usr/local/lib/python3.8/site-packages (from moviepy==1.0.3) (1.23.4)
Collecting imageio<3.0,>=2.5
Downloading https://mirrors.aliyun.com/pypi/packages/97/e2/c5bb16905ab91a0fac03f2a4f1579835bcfea3e297f8cf53e4e2b43c270c/imageio-2.22.2-py3-none-any.whl (3.4 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 3.4/3.4 MB 184.6 kB/s eta 0:00:00
Collecting imageio_ffmpeg>=0.2.0
Downloading https://mirrors.aliyun.com/pypi/packages/e5/3b/fdf3e75462e93b7806ffecad6c5aa35f2cc76b9f2faaedf5e43194ceff09/imageio_ffmpeg-0.4.7-py3-none-manylinux2010_x86_64.whl (26.9 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 26.9/26.9 MB 177.0 kB/s eta 0:00:00
Collecting pillow>=8.3.2
Downloading https://mirrors.aliyun.com/pypi/packages/d8/80/ff6b6ae88982f73d050907dc2c307f387f6a04ce2ca7230ef3a568fbccac/Pillow-9.2.0-cp38-cp38-manylinux_2_28_x86_64.whl (3.2 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 3.2/3.2 MB 180.3 kB/s eta 0:00:00
Collecting idna<4,>=2.5
Downloading https://mirrors.aliyun.com/pypi/packages/fc/34/3030de6f1370931b9dbb4dad48f6ab1015ab1d32447850b9fc94e60097be/idna-3.4-py3-none-any.whl (61 kB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 61.5/61.5 KB 158.4 kB/s eta 0:00:00
Collecting urllib3<1.27,>=1.21.1
Downloading https://mirrors.aliyun.com/pypi/packages/6f/de/5be2e3eed8426f871b170663333a0f627fc2924cc386cd41be065e7ea870/urllib3-1.26.12-py2.py3-none-any.whl (140 kB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 140.4/140.4 KB 194.1 kB/s eta 0:00:00
Collecting certifi>=2017.4.17
Downloading https://mirrors.aliyun.com/pypi/packages/1d/38/fa96a426e0c0e68aabc68e896584b83ad1eec779265a028e156ce509630e/certifi-2022.9.24-py3-none-any.whl (161 kB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 161.1/161.1 KB 184.5 kB/s eta 0:00:00
Collecting charset-normalizer<3,>=2
Downloading https://mirrors.aliyun.com/pypi/packages/db/51/a507c856293ab05cdc1db77ff4bc1268ddd39f29e7dc4919aa497f0adbec/charset_normalizer-2.1.1-py3-none-any.whl (39 kB)
Building wheels for collected packages: moviepy
Building wheel for moviepy (setup.py): started
Building wheel for moviepy (setup.py): finished with status 'done'
Created wheel for moviepy: filename=moviepy-1.0.3-py3-none-any.whl size=110743 sha256=26a5092d88a9ce5490d3fe9177dfbb78aec86fcd55bb206de0d3d35d47a787de
Stored in directory: /root/.cache/pip/wheels/fb/56/20/ea8039f0f19ff16db00131559d9290ff7cddf4711b64a3ff27
Successfully built moviepy
Installing collected packages: urllib3, tqdm, pillow, imageio_ffmpeg, idna, decorator, charset-normalizer, certifi, requests, proglog, imageio, moviepy
Successfully installed certifi-2022.9.24 charset-normalizer-2.1.1 decorator-4.4.2 idna-3.4 imageio-2.22.2 imageio_ffmpeg-0.4.7 moviepy-1.0.3 pillow-9.2.0 proglog-0.1.10 requests-2.28.1 tqdm-4.64.1 urllib3-1.26.12
$ docker images
REPOSITORY TAG IMAGE ID CREATED SIZE
python_media 3.8 755ea9005f1c 10 minutes ago 1.24GB
python 3.8 5e51aed29a27 2 days ago 913MB
4. 创建 docker-compose.yml
$ cd /home/docker/python_media
$ vim docker-compose.yml
version: "3"
services:
python_media:
image: python_media:3.8
container_name: python_media-3.8
restart: always
volumes:
- /home/docker/python_media/work:/home/docker/python_media/work
entrypoint: "sleep infinity"
注：当容器启动后没有 service 运行，但又想继续保持 running 状态，可以添加 entrypoint: "tail -f /dev/null" （即查看空设备的日志，一直会处于阻塞状态），也可以使用 entrypoint: "sleep infinity" （睡眠阻塞）。
5. 启动
$ cd /home/docker/python_media # 进入 docker-compose.yml 所在目录
$ docker-compose up -d # 在后台运行
[+] Running 2/2
⠿ Network python_opencv_default Created 0.0s
⠿ Container python_media-3.8 Started 0.3s
$ docker ps
CONTAINER ID IMAGE COMMAND PORTS NAMES
627ad779b82a python_media:3.8 "sleep infinity" python_media-3.8
6. Python 测试程序
$ cd /home/docker/python_media/work
$ mkdir scripts && cd scripts
$ vim hello.py
#!/usr/bin/python3
# -*- coding: UTF-8 -*-
print("Hello world - 你好!")
$ docker exec -t python_media-3.8 python /home/docker/python_media/work/scripts/hello.py
Hello world - 你好！
7. OpenCV 程序
OpenCV 的几个基础模块:
(1) core 模块: 实现了最核心的数据结构及其基本运算，如绘图函数、数组操作相关函数等；
(2) highgui 模块：实现了视频与图像的读取、显示、存储等接口；
(3) imgproc 模块：实现了图像处理的基础方法，包括图像滤波、图像的几何变换、平滑、阈值分割、形态学处理、边缘检测、目标检测、运动分析和对象跟踪等；
图像处理其它应用的模块：
(1) features2d 模块：用于提取图像特征以及特征匹配；
(2) nonfree 模块：实现了一些专利算法，如 sift 特征；
(3) objdetect 模块：实现了一些目标检测的功能，经典的基于 Haar、LBP 特征的人脸检测，基于 HOG 的行人、汽车等目标检测，分类器使用 Cascade Classification（级联分类）和 Latent SVM 等；
(4) stitching 模块：实现了图像拼接功能；
(5) FLANN（Fast Library for Approximate Nearest Neighbors）模块，包含快速近似最近邻搜索 FLANN 和聚类 Clustering 算法；
(6) ml 模块：机器学习模块（SVM，决策树，Boosting 等）；
(7) photo 模块：包含图像修复和图像去噪两部分；
(8) video 模块：针对视频处理，如背景分离，前景检测、对象跟踪等；
(9) calib3d （即 Calibration 3D）模块：主要是相机校准和三维重建相关的内容。包含了基本的多视角几何算法，单个立体摄像头标定，物体姿态估计，立体相似性算法，3D 信息的重建等等；
(10) G-API 模块：包含超高效的图像处理 pipeline 引擎。
这里演示一个把长视频切割成固定时间的短视频的实例。
1) 视频资源
文件名：test.mp4
文件大小：9.52 MB
时长：14.93 秒
视频源文件：/home/docker/python_media/work/res/test.mp4
2) 创建脚本
$ cd /home/docker/python_media/work/scripts
$ vim opencv_split.py
1 #!/usr/bin/python3 2 3 import os, sys, time 4 import os.path as osp 5 6 import cv2 7 8 9 # 默认值 10 save_dir = '/out' 11 default_seconds = 1 12 max_count = 999 13 14 def splitFile(filename, dura, sdir): 15 16 # Check parameters and file path 17 if filename.isspace() == True or dura < default_seconds or sdir.isspace == True: 18 print("Error: invalid parameters") 19 return 20 21 if osp.exists(filename) != True: 22 print(f"Error: \'{filename}\' not exists") 23 return 24 25 if osp.exists(sdir) != True: 26 os.mkdir(sdir) 27 28 # Open video file and get video parameters 29 print(f"Info: opening \'{filename}\' ...") 30 video_cap = cv2.VideoCapture(filename) 31 32 video_fps = int(video_cap.get(cv2.CAP_PROP_FPS)) 33 video_frames = int(video_cap.get(cv2.CAP_PROP_FRAME_COUNT)) 34 35 duration_end = float(format(video_frames/video_fps, '.2f')) 36 if duration_end <= dura: 37 print(f"Error: video duration {duration_end} seconds <= {dura} seconds, unable to split") 38 return 39 40 # Calculate split file count 41 filecount = int(duration_end // dura) 42 if (duration_end % dura) > 1.0: 43 filecount += 1 44 45 if filecount > max_count: 46 print(f"Error: file count {filecount} > {max_count}, unable to split") 47 return 48 49 # 50 print(f"Info: split to {filecount} files ...") 51 52 frame_index = 1 53 file_index = 1 54 55 fileObj = osp.splitext(osp.basename(filename)) 56 output_fullname = "{0}/{1}_{2:03d}{3}".format(sdir, fileObj[0], file_index, fileObj[1]) 57 58 dura_frames = dura * video_fps 59 frame_size = (int(video_cap.get(cv2.CAP_PROP_FRAME_WIDTH)), int(video_cap.get(cv2.CAP_PROP_FRAME_HEIGHT))) 60 61 fourcc = cv2.VideoWriter_fourcc('m','p','4','v') 62 video_writer = cv2.VideoWriter(output_fullname, fourcc, video_fps, frame_size) 63 64 print(f"Info: writing \'{output_fullname}\' - {file_index}/{filecount}\n... ") 65 66 while True: 67 68 # Read a frame 69 is_success, bgr_im = video_cap.read() 70 71 # Break when no more frame 72 if not is_success: 73 video_writer.release() 74 video_cap.release() 75 break 76 77 # Write current frame to output file 78 if ((frame_index % dura_frames) != 0): 79 if video_writer.isOpened(): 80 video_writer.write(bgr_im) 81 82 # Final frame of the current output file: 83 # 1. write the final frame , 2. close output file， 3. create next output file 84 if ((frame_index % dura_frames) == 0): 85 if video_writer.isOpened(): 86 video_writer.write(bgr_im) 87 video_writer.release() 88 89 file_index += 1 90 output_fullname = "{0}/{1}_{2:03d}{3}".format(sdir, fileObj[0], file_index, fileObj[1]) 91 video_writer = cv2.VideoWriter(output_fullname, fourcc, video_fps, frame_size) 92 93 print(f"Info: writing \'{output_fullname}\' - {file_index}/{filecount}\n... ") 94 95 frame_index += 1 96 97 video_cap.release() 98 99 100 # 命令行入口 101 if __name__=="__main__": 102 msg = """Usage: python %s [file] [seconds] [sdir] 103 file - Source file, full path 104 seconds - Duration, default 10 seconds 105 sdir - Output dir 106 """ 107 count = len(sys.argv) 108 if count < 2: 109 print(msg % sys.argv[0]) 110 elif count == 2: 111 dura = default_seconds 112 sdir = osp.abspath('.') + save_dir 113 splitFile(sys.argv[1], dura, sdir) 114 elif count == 3: 115 dura = int(sys.argv[2]) 116 if dura < default_seconds: 117 dura = default_seconds 118 sdir = osp.abspath('.') + save_dir 119 splitFile(sys.argv[1], dura, sdir) 120 else: 121 dura = int(sys.argv[2]) 122 if dura < default_seconds: 123 dura = default_seconds 124 sdir = sys.argv[3] 125 if sdir.isspace() == True: 126 sdir = osp.abspath('.') + save_dir 127 splitFile(sys.argv[1], dura, sdir)
3) 运行
$ docker exec -t python_media-3.8 python /home/docker/python_media/work/scripts/opencv_split.py /home/docker/python_media/work/res/test.mp4 5
Info: opening '/home/docker/python_media/work/res/test.mp4' ...
Info: split to 3 files ...
Info: writing '/home/docker/python_media/work/out/test_001.mp4' - 1/3
...
Info: writing '/home/docker/python_media/work/out/test_002.mp4' - 2/3
...
Info: writing '/home/docker/python_media/work/out/test_003.mp4' - 3/3
...
分割后的小视频在 /home/docker/python_media/work/out 目录，可以正常播放，但没有音频（声音），test_001.mp4 ~ test_002.mp4 时长都是 5 秒，test_003.mp4 时长 4 秒。
注：以上程序分割的小视频，没有音频（声音），而且生成的文件比源文件还大，这是因为输出文件时没有设置比特率。本文只演示如何分割文件，Opencv 的音频和参数的设置在以后关于 Opencv 的文章里再讲解。简单的文件分割，建议使用 Moviepy 来实现。
8. Moviepy 程序
1) 创建脚本
$ cd /home/docker/python_media/work/scripts
$ vim moviepy_split.py
1 #!/usr/bin/python3 2 3 import os, sys 4 import os.path as osp 5 6 from moviepy import editor 7 8 # 默认值 9 save_dir = '/out2' 10 default_seconds = 1 11 max_count = 999 12 13 # 分割函数 14 def splitFile(filename, dura, sdir): 15 if filename.isspace() == True or dura < default_seconds or sdir.isspace == True: 16 print("Error: invalid parameters") 17 return 18 19 if osp.exists(filename) != True: 20 print(f"Error: \'{filename}\' not exists") 21 return 22 23 if osp.exists(sdir) != True: 24 os.mkdir(sdir) 25 26 print(f"Info: opening \'{filename}\' ...") 27 video_clip = editor.VideoFileClip(filename) 28 duration_end = video_clip.duration 29 if duration_end <= dura: 30 print(f"Error: video duration {video_clip.duration} seconds <= {dura} seconds, unable to split") 31 return 32 33 filecount = int(duration_end // dura) 34 if (duration_end % dura) > 1.0: 35 filecount += 1 36 37 if filecount > max_count: 38 print(f"Error: file count {filecount} > {max_count}, unable to split") 39 return 40 41 print(f"Info: split to {filecount} files ...") 42 # 43 fileObj = osp.splitext(osp.basename(filename)) 44 45 i = 1 46 start = 0 47 end = dura 48 while i <= filecount: 49 50 if start > duration_end: 51 break 52 53 if end > duration_end: 54 end = duration_end 55 56 sub_clip = video_clip.subclip(start, end) 57 sub_clip.write_videofile("{0}/{1}_{2:03d}{3}".format(sdir, fileObj[0], i, fileObj[1])) 58 59 print(f"Info: in progress -> {i} / {filecount}\n") 60 61 start += dura 62 end += dura 63 i += 1 64 65 66 # 命令行入口 67 if __name__=="__main__": 68 msg = """Usage: python %s [file] [seconds] [sdir] 69 file - Source file, full path 70 seconds - Duration, default 10 seconds 71 sdir - Output dir 72 """ 73 count = len(sys.argv) 74 if count < 2: 75 print(msg % sys.argv[0]) 76 elif count == 2: 77 dura = default_seconds 78 sdir = osp.abspath('.') + save_dir 79 splitFile(sys.argv[1], dura, sdir) 80 elif count == 3: 81 dura = int(sys.argv[2]) 82 if dura < default_seconds: 83 dura = default_seconds 84 sdir = osp.abspath('.') + save_dir 85 splitFile(sys.argv[1], dura, sdir) 86 else: 87 dura = int(sys.argv[2]) 88 if dura < default_seconds: 89 dura = default_seconds 90 sdir = sys.argv[3] 91 if sdir.isspace() == True: 92 sdir = osp.abspath('.') + save_dir 93 splitFile(sys.argv[1], dura, sdir)
2) 运行
$ docker exec -t python_media-3.8 python /home/docker/python_media/work/scripts/moviepy_split.py /home/docker/python_media/work/res/test.mp4 5
Info: opening '/home/docker/python_media/work/res/test.mp4' ...
Info: split to 3 files ...
Moviepy - Building video /home/docker/python_media/work/out2/test_001.mp4.
MoviePy - Writing audio in test_001TEMP_MPY_wvf_snd.mp3
MoviePy - Done.
Moviepy - Writing video /home/docker/python_media/work/out2/test_001.mp4
Moviepy - Done !
Moviepy - video ready /home/docker/python_media/work/out2/test_001.mp4
Info: in progress -> 1 / 3
Moviepy - Building video /home/docker/python_media/work/out2/test_002.mp4.
MoviePy - Writing audio in test_002TEMP_MPY_wvf_snd.mp3
MoviePy - Done.
Moviepy - Writing video /home/docker/python_media/work/out2/test_002.mp4
Moviepy - Done !
Moviepy - video ready /home/docker/python_media/work/out2/test_002.mp4
Info: in progress -> 2 / 3
Moviepy - Building video /home/docker/python_media/work/out2/test_003.mp4.
MoviePy - Writing audio in test_003TEMP_MPY_wvf_snd.mp3
MoviePy - Done.
Moviepy - Writing video /home/docker/python_media/work/out2/test_003.mp4
Moviepy - Done !
Moviepy - video ready /home/docker/python_media/work/out2/test_003.mp4
Info: in progress -> 3 / 3
分割后的小视频在 /home/docker/python_media/work/out2 目录，可以正常播放，音频（声音）正常，test_001.mp4 ~ test_002.mp4 时长都是 5 秒，test_003.mp4 时长 4 秒。
浙公网安备 33010602011771号