YouTube Live Webcam Human Detector

This project watches a YouTube live stream (webcam) and detects humans in real-time, drawing bounding boxes around them and saving the processed frames as images. It also provides a web interface to view the processed stream in real-time.

I'm just doing this for fun. This is not for any other use.

Screenshot

Installation

Option 1: Local Installation

1. Clone this repository
2. Install the required packages:

pip install -r requirements.txt

3. Make sure you have FFmpeg installed on your system
4. Set up your environment configuration:

cp .env.example .env

5. Edit the .env file with your preferred settings:

YOUTUBE_URL=https://www.youtube.com/watch?v=your_youtube_video_id

Option 2: Docker Installation

1. Make sure you have Docker and Docker Compose installed on your system
2. Set up your environment configuration:

cp .env.example .env

3. Edit the .env file with your YouTube URL
4. Build and run the container:

docker compose up --build

Usage

Local Usage

Run the script:

python live_human_detector.py

The script will automatically:

• Connect to the YouTube stream specified in your .env file
• Process the video frames
• Detect and mark humans with bounding boxes
• Save the processed video to the output folder
• Start a web server at http://localhost:5005

Docker Usage

Start the container:

docker compose up

To stop the container:

docker compose down

Viewing the Live Stream

Once the application is running, open a web browser and navigate to:

http://localhost:5005

You'll see the live stream with bounding boxes drawn around detected humans.

Configuration

The following environment variables can be set in the .env file:

Variable	Description	Default
YOUTUBE_URL	URL of the YouTube live stream to process	https://www.youtube.com/watch?v=your_youtube_video_id
CONFIDENCE_THRESHOLD	Minimum confidence level (0-1) for human detection	0.5
SAVE_OUTPUT	Whether to save the processed video to disk	true

Output

The processed frames are saved to timestamped directories in the output folder:

output/session_20230815_143027/
  ├── frame_000001.jpg
  ├── frame_000002.jpg
  ├── frame_000003.jpg
  └── ...

Frames are saved in the following cases:

• When humans are detected in the frame
• Periodically (every 5 seconds) to maintain context

Each saved frame includes:

• Bounding boxes around detected humans
• The current FPS calculation
• A timestamp showing when the frame was captured

Requirements

• Python 3.7+
• OpenCV
• PyTorch
• yt-dlp (YouTube downloader)
• FFmpeg (for video processing)
• Flask (for web interface)

Docker Requirements

• Docker
• Docker Compose

Note

This application uses a pre-trained Faster R-CNN model from PyTorch for human detection. The model is optimized for detecting people in various contexts.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Disclaimer

IMPORTANT: This software is provided for educational and research purposes only.

• This project is designed for analyzing publicly available streams only.
• Always respect privacy and copyright laws in your jurisdiction.
• Do not use this software for surveillance or to violate anyone's privacy.
• The authors and contributors are not responsible for any misuse of this software.