This project was developed during my Research & Development Internship at HILCPS Lab, IIT Madras, in collaboration with FedEx. It focuses on real-time driver monitoring, vehicle health tracking, and fleet analytics using Flutter, Computer Vision, ML, and Embedded Systems.
The system consists of:
- **Flutter App (Mobile) ** – Real-time driver monitoring, vehicle diagnostics, and alerts.
- **Admin Dashboard (Web) ** – Fleet tracking, weekly/monthly reports, and AI-driven insights.
- **Backend (WebSocket & PostgreSQL) ** – Data ingestion, real-time storage, and analytics.
- Programming: Python, Dart, SQL
- Machine Learning & Computer Vision: OpenCV, Dlib, MediaPipe, NumPy
- Deep Learning Models: CNN, RNN, LSTM
- NLP & AI Techniques: Gaze Tracking, Facial Landmark Detection, Pose Estimation
- Frameworks & Tools: Flutter, Flask, PostgreSQL, WebSockets, ELM327 (OBD-II), Docker
- Drowsiness Detection using:
- Eye Aspect Ratio (EAR) – Detects eye closure duration.
- Mouth Aspect Ratio (MAR) – Identifies yawning frequency.
- Distraction Detection using:
- Yaw, Pitch, Roll – Measures head movement angles.
- Gaze Points Tracking – Determines if the driver is looking away.
- Combined these features to provide real-time alerts with audio & emojis 🚨🎵😴.
- Used ELM327 OBD-II scanner to retrieve vehicle diagnostics:
- Speed, RPM, Fuel Pressure, Engine Oil Temperature
- Simulated ELM327 data due to lack of car access, helping understand CAN protocol.
- Integrated vehicle & driver parameters for real-time fleet analytics.
- Live tracking & analytics (location, alerts, driving behavior).
- Weekly & Monthly Reports on driver performance & vehicle health.
- WebSocket communication for seamless real-time data updates.
| Challenge | Solution |
|---|---|
| Limited car access for ELM327 testing | Built a simulator to test OBD-II data flow. |
| Optimizing battery usage in the Flutter app | Implemented background processing for efficient data collection. |
| Ensuring accurate distraction detection | Combined multiple vision-based parameters for higher precision. |
| Real-time alert system without user annoyance | Used cool emojis & subtle audio alerts to improve UX. |
Enhanced driver safety with AI-driven alerts.
Improved vehicle performance monitoring using real-time analytics.
Fleet-wide optimization for data-driven decision-making.
This project represents a fusion of AI, embedded systems, and real-time analytics to improve fleet safety and efficiency.
For collaboration or inquiries, feel free to connect!