restaurant-ai

A complete Python-based AI system for smart restaurant analytics using computer vision. The system uses video feed (CCTV or recorded video) to detect and track human footfall, analyze staff activity, and provide actionable insights.

Stars
6
Language
Python
Updated
View on GitHub

Restaurant AI - Smart Restaurant Analytics

A complete Python-based AI system for smart restaurant analytics using computer vision. The system uses video feed (CCTV or recorded video) to detect and track human footfall, analyze staff activity, and provide actionable insights.

Features

  • People Detection: YOLOv8-based person detection
  • Object Tracking: DeepSORT-style tracking with unique IDs
  • Footfall Counting: Virtual entry/exit line counting
  • Zone-Based Tracking: Track time spent in different zones
  • Wait Time Estimation: Measure customer wait times
  • Staff Detection: Color-based staff uniform detection
  • Staff Efficiency Metrics: Idle time, movement frequency, interactions
  • Analytics Engine: Generate insights on staffing and service quality
  • Visualization: Interactive plots and heatmaps

Project Structure

restaurant-ai/
├── notebooks/
│   ├── detection.ipynb      # YOLO person detection
│   ├── tracking.ipynb        # Object tracking
│   ├── analytics.ipynb      # Analytics and insights
│   └── visualization.ipynb  # Data visualization
├── models/                  # YOLO models
├── utils/                   # Utility modules
│   ├── detection.py         # Person detection
│   ├── tracking.py         # Object tracking
│   ├── analytics.py        # Analytics components
│   └── visualization.py    # Plotting functions
├── data/                    # Video input
├── outputs/                 # Analytics output
├── app/                     # FastAPI backend
└── requirements.txt        # Dependencies

Installation

  1. Clone the repository:
git clone https://github.com/Manavarya09/restaurant-ai.git
cd restaurant-ai
  1. Create and activate virtual environment:
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate
  1. Install dependencies:
pip install -r requirements.txt
  1. Download YOLOv8 model (will be auto-downloaded on first run):
yolo export model=yolov8n.pt  # Optional: download model explicitly

Quick Start

Option 1: Use Sample Video

  1. Download a sample video:
# Download sample video from common sources
# Example using wget (replace with actual video URL)
wget -O data/sample_video.mp4 "https://example.com/sample.mp4"

Or use your own video file by placing it in data/sample_video.mp4.

Option 2: Use Camera

The system can use webcam or CCTV camera as input. Just change the video path in notebooks.

Running the Notebooks

  1. Start Jupyter:
jupyter notebook
  1. Open and run notebooks in order:
    1. detection.ipynb - Test person detection
    2. tracking.ipynb - Test object tracking
    3. analytics.ipynb - Run complete analytics
    • visualization.ipynb - View results

FastAPI Backend

Start the API server:

cd app
uvicorn main:app --reload

API Endpoints:

  • GET / - API documentation
  • GET /analytics - Get current analytics
  • POST /update - Update analytics data
  • POST /video/analyze - Analyze video frame
  • GET /health - Health check
  • POST /reset - Reset analytics

Configuration

Zone Definition

Edit zone coordinates in analytics.ipynb:

zone_manager.add_zone('entrance', [(x1, y1), (x2, y2), ...])
zone_manager.add_zone('waiting', [(x1, y1), (x2, y2), ...])
zone_manager.add_zone('dining', [(x1, y1), (x2, y2), ...])

Staff Detection

Adjust HSV thresholds for staff uniform:

staff_detector = StaffDetector(
    hsv_lower=(0, 0, 0),      # Lower bound
    hsv_upper=(180, 50, 80)   # Upper bound
)

Entry/Exit Line

Set counting line position:

footfall_counter = FootfallCounter(
    line_start=(x1, y1),
    line_end=(x2, y2)
)

Sample Analytics Output

==================================================
CURRENT STATISTICS
==================================================
Entries: 45
Exits: 38
Current customers: 7
Staff count: 3
Average wait time: 180.5s (3.0 min)
Average idle time: 45.2s
==================================================

INSIGHTS AND RECOMMENDATIONS
==================================================

Recommendations:
  - Customer-to-staff ratio is 2.3:1 - consider reducing staff

Requirements

  • Python 3.10+
  • OpenCV
  • Ultralytics YOLOv8
  • NumPy, Pandas
  • Matplotlib, Plotly
  • FastAPI (for backend)

Troubleshooting

CUDA/GPU Issues

For GPU acceleration, install torch with CUDA:

pip install torch torchvision --index-url https://download.pytorch.org/whl/cu118

Mac (Apple Silicon) Issues

Install compatible packages:

pip install opencv-python numpy pandas
pip install torch torchvision --index-url https://download.pytorch.org/whl/cpu

Video Not Found

If video file is not found, the system will automatically use camera (webcam) as fallback.

License

MIT License

Author

Manavarya09