SENTRA is an AI-driven proactive safety platform designed to help women travel safely by combining intelligent routing, real-time monitoring, and on-device threat detection.
Unlike traditional safety apps that depend only on a manual SOS button, SENTRA continuously evaluates the surrounding environment and guides users through safer routes while monitoring potential threats automatically.
The system integrates AI models, geospatial analysis, and sensor data to detect risks and notify authorities in real time.
Traditional navigation apps optimize for shortest distance or fastest time. SENTRA introduces safety-aware routing.
The system evaluates multiple routes and assigns a safety score based on environmental factors.
- Base Score: Distance of route (shorter routes preferred)
- Safety Bonus: Segments near open businesses or crowded areas
- Risk Penalty: Segments passing through high-risk zones
The safest route is highlighted for the user.
Continuous monitoring can drain device battery. SENTRA solves this using zone-based adaptive polling.
| Zone Type | Behavior |
|---|---|
| Green Zone | GPS updates every 5 minutes, monitoring disabled |
| Yellow Zone | GPS updates every 1 minute, sensors on standby |
| Red Zone | Real-time GPS streaming and active threat detection |
This ensures maximum safety with minimal battery usage.
SENTRA includes an on-device machine learning model that detects potential threats using environmental audio.
The model detects sounds such as:
- Screaming
- Glass breaking
- Aggressive shouting
To reduce false alarms, SENTRA verifies multiple signals before triggering alerts.
if detect_scream(confidence > 0.85):
if motion == "running" or motion == "violent_shaking":
trigger_sos()
else:
vibrate_phone_warning()All audio processing occurs locally on the device, ensuring user privacy.
Instead of manually marking unsafe areas, SENTRA automatically identifies risk zones using clustering algorithms.
DBSCAN (Density Based Spatial Clustering)
- Historical incident reports
- Crowd density indicators
- Location patterns
Automatically generated high-risk zones displayed on the map.
| Component | Technology | Responsibility |
|---|---|---|
| Mobile App | Flutter / React Native | UI, sensor data collection, TFLite inference |
| Backend API | Python (FastAPI / Flask) | Route scoring logic, zone management |
| Database | Firebase / Supabase | Authentication, location sync, zone storage |
| ML Engine | TensorFlow Lite | On-device audio threat detection |
| Dashboard | React.js / Next.js | Monitoring alerts and incident visualization |
- Flutter / React Native
- Python FastAPI
- TensorFlow Lite
- Firebase / Supabase
- Mapbox / Google Maps API
- DBSCAN Clustering Algorithm
sentra/
│
├── mobile-app/
│ ├── ui/
│ ├── sensors/
│ └── navigation/
│
├── backend/
│ ├── api/
│ ├── routing-engine/
│ └── zone-clustering/
│
├── ml-models/
│ ├── audio-detection/
│ └── training-scripts/
│
├── dashboard/
│ ├── alerts/
│ └── heatmaps/
│
└── datasets/
git clone https://github.com/yourusername/sentra.git
cd sentracd backend
pip install -r requirements.txt
uvicorn main:app --reloadcd mobile-app
flutter pub get
flutter run-
User starts navigation in the app.
-
SENTRA fetches multiple possible routes.
-
Each route is analyzed using safety scoring logic.
-
The safest route is suggested.
-
While travelling:
- location updates are monitored
- sensors activate in high-risk zones
-
If a threat is detected:
- SOS alert is triggered
- location is sent to the monitoring dashboard.
- All audio analysis is performed on-device
- No continuous recording or cloud storage of user audio
- Location data is encrypted and shared only during emergencies.
- Real crime dataset integration
- Crowd-sourced incident reporting
- Smart wearable integration
- AI-based video anomaly detection
- Predictive safety analytics using historical patterns
- Architecture diagram
- App UI screenshots
- Animated system flow diagram