Automatic YouTube Visual Data Collection

Make your life easy :)

Note: FFMPEG is required for OpenCV to work with local downloaded videos (especially .AVI files) make sure you download and use latest FFMPEG from this link.

Build your own dataset for object recognition and object detection/segmentation

You can use this python code to download the frames of youtube video without downloading video.

Step 1. pip install -r requirements.txt

Then select the output directory where you want to save the frames (edit python code to set paths and directories)

Step 2. update the .py file

Add youtube video and output dir in python script

step 3. python <script-name.py>

It will start capturing frames from youtube video and download it into the specified folders.

Usage

Alternatively you can use the cli.py file if you do not want to manually modify the main script.

>>> python cli.py --help
>>> usage: downlaod_youtube_frames. [-h] --videolink VIDEOLINK --destination DESTINATION

  Program to automatically download youtube images datasets.

  optional arguments:
    -h, --help            show this help message and exit
    --videolink VIDEOLINK
                         YouTube video link
    --destination DESTINATION
                         Target path to save imgz

  A simple and nice cli script to create youtube datasets

How to run on cli/terminal?

Example run

>>> python cli.py --videolink "https://www.youtube.com/watch?v=PWRg_wak9oI" --destination "D:\dataset\test"

Running like above example will start downloading the video frames into the provided destination. Optionally, you can provide --showframe argument in command line if you want to show the frames that are being saved to local disk.

Example

>>> python cli.py --videolink "https://www.youtube.com/watch?v=PWRg_wak9oI" --destination "D:\dataset\test" --showframe

NOTE: Make sure --videolink and --destination are both strings (enclosed in quotation marks)

Todo

• Add framecount in argparse
• work with any online videos ( YouTube + more )

In Progress

• Working on improving readme, How-To and upload on PyPI.

Done ✓

• Capturing frames from remote YouTube videos.
• merge this to deepds
• Saving frames to target directory.
• Cli.py file for easy usage.