Still working on the project...

To run the project:

1. Clone the repository

git clone https://github.com/KumaarBalbir/droneAnalyzer.git

2. Create a virtual environment in the root directory of the project On Ubuntu or Debian based systems:

python3 -m venv <venv_name>

On Windows:

python -m venv <venv_name>

3. Activate your virtual environment: On Ubuntu or Debian based systems:

source <venv_name>/bin/activate

On Windows:

<venv_name>\Scripts\activate

3. Install uv (a fast Python package manager that acts as a drop-in replacement for pip)

pip install uv

4. Install the project dependencies

uv pip install -r pyproject.toml

3. Run the project

streamlit run main.py --server.port 5000

4. View the project in your browser

http://localhost:5000

5. The project architecture:

Design Decisions & Architecture

Core Architecture

1. Modular Design

   • Separated concerns into distinct modules (processors, models, storage, alerts)
   • Each module has a single responsibility
   • Easy to extend and maintain

2. Data Flow Pipeline

   Video Input → Frame Processing → Object Detection → Event Analysis → Storage/Alerts
   

   • Linear pipeline for predictable data flow
   • Each stage processes data independently
   • Easy to add new processing stages

3. Vision Language Model Choice (BLIP)

   • Selected over CLIP for better scene understanding
   • Provides detailed natural language descriptions
   • More suitable for security surveillance use case
   • Better at describing activities and relationships

4. Storage Design (will shift from in Memory to persistent storage)

   • Frame-by-frame indexing for historical analysis
   • Separate event logging for security incidents
   • Enables efficient querying by time and object type
   • Supports both real-time and historical analysis

Technical Choices

1. Video Processing

   • OpenCV for efficient frame capture and processing
   • Configurable frame sampling rate (1 FPS default)
   • Support for both live streams and recorded videos
   • Memory-efficient processing

2. Object Detection

   • YOLOv8 for real-time object detection
   • Configurable confidence thresholds
   • Supports multiple object classes
   • Efficient inference on CPU/GPU

3. Data Structures

   • Standardized frame data format
   • Consistent event logging structure
   • Type hints for better code maintainability
   • JSON-compatible data structures

4. API Design

   • Simple, intuitive interfaces
   • Consistent method naming
   • Clear input/output contracts
   • Easy to integrate with other systems

Performance Considerations

1. Scalability
   • Modular design allows horizontal scaling
   • Efficient indexing for quick queries

Security Features

1. Alert System
   • Configurable security rules
   • Multiple alert severity levels
   • Real-time alert generation
   • Historical alert tracking

Future Extensibility

1. Planned Enhancements

   • Additional ML model support
   • Enhanced query capabilities
   • Advanced analytics features
   • Integration with external systems

2. Integration Points

   • Shift to persistent frame indexing
   • Standard API endpoints
   • Webhook support
   • Database adapters
   • Message queue integration

Documentation

The full documentation is available at: https://KumaarBalbir.github.io/droneAnalyzer/

For local development:

cd docs
make html
# Open docs/build/html/index.html in your browser