Camera Streamer

A Python-based camera streaming and recording system that can connect to a camera, stream video in real-time, and automatically save video segments at regular intervals.

Quick Start with GUI

Easiest way to use all features: Launch the graphical control panel:

python camera_control_panel.py

This opens a settings window where you can:

• Select which mode to run (streaming, detection, recording, etc.)
• Configure all settings (camera, model, confidence, output directory, etc.)
• Start/stop the camera with one click
• No need to remember command-line arguments!

Features

Stage 1: Basic Camera Streaming

• Connect to camera devices
• Real-time video streaming with frame counter
• Display camera properties (resolution, FPS)

Stage 2: Video Recording System

• Continuous video recording from camera
• Automatic segmentation (saves video every minute by default)
• Timestamped filenames for easy organization
• Preview window with recording status
• Configurable segment duration and output directory

Person Detection with YOLO

• Real-time person detection using YOLOv8
• Draws bounding boxes around detected persons
• Shows detection confidence scores
• Multiple model sizes for speed/accuracy tradeoff
• Configurable confidence threshold

Person-Activated Recording

• Continuously monitors camera for persons
• Automatically starts recording when person enters frame
• Stops recording when person leaves frame
• Saves video segments only when persons are detected
• Prevents flickering with configurable grace period

Person with Cup Detection

• Detects persons holding cups/mugs
• Only records when a person with a cup is detected
• Ignores persons without cups
• Uses spatial proximity to determine if person is holding cup
• Saves recordings to separate directory

Installation

1. Install Python 3.7 or higher

2. Install required dependencies:

pip install -r requirements.txt

Note: The GUI control panel uses tkinter, which is included with Python on most systems. If you get an error about tkinter, install it:

• Ubuntu/Debian: sudo apt-get install python3-tk
• macOS: Usually pre-installed
• Windows: Usually pre-installed

Usage

GUI Control Panel (Recommended)

The easiest way to use all features is through the graphical control panel:

python camera_control_panel.py

Features:

• Mode Selection: Choose from all available modes
  • Basic Streaming
  • Person Detection (view only)
  • Person Recording
  • Person with Cup Recording
• Settings Configuration: Adjust all parameters through GUI
  • Camera index
  • YOLO model size and confidence
  • Output directory
  • Recording duration and grace periods
• One-Click Control: Start/Stop buttons
• Real-time Status: See current status of camera operations

Command Line Usage

For advanced users or scripting, you can also use the command-line interface:

Stage 1: Basic Streaming

To stream video from your camera:

python camera_stream.py

Or specify a different camera index:

python camera_stream.py --camera 1

Press 'q' to quit the stream.

Stage 2: Video Recording

To record video segments (saves every 60 seconds by default):

python camera_recorder.py

Command-line options:

• --camera N: Specify camera index (default: 0)
• --duration N: Segment duration in seconds (default: 60)
• --output DIR: Output directory for video files (default: recordings)
• --no-preview: Disable preview window (useful for headless operation)

Examples:

Record with 30-second segments:

python camera_recorder.py --duration 30

Record to a custom directory:

python camera_recorder.py --output /path/to/videos

Record without preview window:

python camera_recorder.py --no-preview

Person Detection

To detect persons in the camera stream with bounding boxes:

python person_detector.py

Command-line options:

• --camera N: Specify camera index (default: 0)
• --model SIZE: YOLO model size - nano (fastest), small, medium, large, xlarge (most accurate, default: nano)
• --confidence FLOAT: Confidence threshold 0.0-1.0 (default: 0.25)

Examples:

Use a larger model for better accuracy:

python person_detector.py --model small

Adjust confidence threshold (higher = fewer false positives):

python person_detector.py --confidence 0.5

Use different camera:

python person_detector.py --camera 1

Note: On first run, the YOLO model will be automatically downloaded (this may take a moment).

Person-Activated Recording

Monitor camera and automatically record only when persons are detected:

python person_recorder.py

How it works:

• Camera continuously monitors for persons
• Recording starts automatically when a person enters the frame
• Recording stops when no person is detected (after a short grace period)
• Each recording is saved as a separate video file with timestamp in filename
• Timestamp overlays: Every frame in recorded videos includes a timestamp overlay (top-right corner)
• Recording log: All recording events are logged to recording_log.txt with precise timestamps

Command-line options:

• --camera N: Camera index (default: 0)
• --model SIZE: YOLO model size (default: nano)
• --confidence FLOAT: Confidence threshold 0.0-1.0 (default: 0.25)
• --output DIR: Output directory (default: person_recordings)
• --min-frames N: Minimum frames before stopping (prevents flickering, default: 30)
• --no-preview: Disable preview window

Examples:

Record with higher confidence threshold:

python person_recorder.py --confidence 0.5

Use more accurate model:

python person_recorder.py --model small

Custom output directory:

python person_recorder.py --output /path/to/recordings

Output:

• Video files: Saved with timestamped filenames like:

  • person_20251208_143022.mp4 - Recording when person entered
  • person_20251208_143145.mp4 - Next recording when person entered again

• Timestamp overlays: Each recorded video frame displays:

  • Current date and time (top-right corner)
  • Recording duration (elapsed time since recording started)

• Recording log: recording_log.txt file contains:

  • Timestamp of each recording start
  • Timestamp of each recording stop
  • Duration and frame count for each recording
  • Session summary statistics

Person with Cup Detection

Monitor camera and record only when a person holding a cup is detected:

python person_cup_detector.py

How it works:

• Continuously monitors for persons and cups
• Uses spatial proximity to detect if a person is holding a cup
• Only records when person + cup combination is detected
• Ignores persons without cups
• Saves recordings to separate directory (person_cup_recordings/)

Command-line options:

• --camera N: Camera index (default: 0)
• --model SIZE: YOLO model size (default: nano)
• --confidence FLOAT: Confidence threshold 0.0-1.0 (default: 0.25)
• --output DIR: Output directory (default: person_cup_recordings)
• --min-frames N: Minimum frames before stopping (default: 30)
• --no-preview: Disable preview window

Examples:

# Use more accurate model
python person_cup_detector.py --model small

# Adjust confidence threshold
python person_cup_detector.py --confidence 0.5

# Custom output directory
python person_cup_detector.py --output /path/to/cup_recordings

Visual indicators:

• Blue boxes: Detected persons
• Yellow boxes: Detected cups
• Green boxes: Person with cup (recording trigger)
• Green line: Connection between person and cup

Output:

• Video files saved as: person_cup_YYYYMMDD_HHMMSS.mp4
• Timestamp overlays on each frame
• Recording log file: recording_log.txt

Output

Video segments are saved in the recordings directory (or your specified output directory) with filenames like:

• segment_20240115_143022.mp4
• segment_20240115_143122.mp4
• etc.

Each filename includes a timestamp indicating when the recording started.

Notes

• The system will automatically try alternative camera indices if the default camera (index 0) is not available
• Video files are saved in MP4 format using the 'mp4v' codec
• If the camera FPS cannot be detected, a default of 30 FPS is used
• Press 'q' in the preview window to stop recording gracefully

Troubleshooting

Camera not detected:

• Try different camera indices: --camera 1 or --camera 2
• Check that your camera is properly connected
• On Linux, you may need to check permissions: ls -l /dev/video*

Permission errors:

• On Linux, you may need to add your user to the video group:

  sudo usermod -a -G video $USER

  Then log out and log back in.

Video playback issues:

• If MP4 files don't play, you may need to install additional codecs
• Consider using a different codec by modifying the fourcc parameter in camera_recorder.py