NOVA is an Android application designed to help visually impaired users navigate their surroundings safely. It uses real-time object detection and depth estimation to identify obstacles and announce them through audio feedback.

Android Camera → JPEG Frame → Python Server (YOLO + MiDaS) → Detected Objects + Distances → Text-to-Speech → Audio Output

1. The app captures frames from the phone's camera and sends them to a Python server
2. The server runs YOLOv26 object detection to identify objects in the frame
3. MiDaS depth estimation calculates the distance to each detected object
4. Results are sent back to the app and converted to speech (e.g. "person at 0.2 meters detected") using the ElevenLabs TTS API

NOVA/
├── app/                        # Android application
│   └── app/src/main/java/com/example/impairedapp/
│       └── MainActivity.kt     # Main app logic (camera, networking, TTS)
└── server/                     # Python backend
    ├── server.py               # FastAPI server with YOLO detection
    ├── distanceCalculator.py   # MiDaS depth-to-distance conversion
    ├── yolo.py                 # Standalone YOLO test script
    └── *.pt                    # YOLO model weights (nano, medium, large)

• Android Studio
• Python 3.8+
• Android device with camera (API 24+)

cd server
python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate

pip install fastapi uvicorn ultralytics torch torchvision opencv-python numpy matplotlib

# Create config.py with your server IP
echo 'SERVER_IP = "0.0.0.0"' > config.py

python server.py
# Server runs on http://0.0.0.0:8000

1. Open the app/ directory in Android Studio
2. Update the server URL in MainActivity.kt to point to your server's IP address
3. Add your ElevenLabs API key
4. Build and run on a physical device:

   ./gradlew installDebug

Note: A physical Android device is recommended since the app relies on the camera for real-time object detection.

The app requires:

• Camera - for real-time environment scanning
• Internet - for server communication and TTS API calls

Response format:

{
  "objects": [["person", 0.2], ["car", 3.6]]
}