The script uses yt_dlp to download videos from YouTube playlists and categorize them into folders according to actions like abuse, arrest, robbery, etc.
- Python 3.10 or later
yt_dlplibrary- Cookies from browser to access restricted videos (if needed)
- Conda is recommended for creating virtual environments:
conda create -n yt_downloader python=3.10
conda activate yt_downloader
pip install yt-dlp-
To download restricted videos or videos that require login, you need to use an extension to get cookies from your browser:
-
Extension: Get cookies.txt
-
After installing, go to YouTube, log in with your account, click on the extension to download the cookies file.
-
Save the file (any name is fine, for example
cookies.txt).
- Place the cookies file in the same directory as the script, and update the following line in the Python code if the file name is different:
"cookiefile": "your_cookies_filename.txt"- Then run:
python crawl_data.pyThe video will be uploaded to the crawled data folder, each topic is a subfolder. The video is named in the format:
crawled_{topic_name}_{000001}.mp4
For example:
crawled data/
├— abuse/
│ └— crawled_abuse_000001.mp4
├— robbery/
│ └— crawled_robbery_000001.mp4
...
## Note
* Cookies are **machine specific**, so each person needs to create their own using their browser.
* If you encounter any errors, please check your playlist URL or update `yt_dlp` with the command:
```bash
pip install -U yt-dlp
Collected YouTube videos are categorized into specific action-based folders such as:
abusearrestrobberynormalexplosionfight- ...
Each playlist contains surveillance or incident footage matching the category. This helps structure the dataset for action recognition or anomaly detection.
Categorized YouTube playlists by action
