🚘 SARATHI – Unified Attention Monitoring for Transportation

"A moment’s lapse. A lifetime lost. SARATHI watches, warns, and saves."

---

📍 Overview

SARATHI is a real-time AI-based driver and pilot attention monitoring system that detects drowsiness, unsafe posture, and inattentiveness. It serves as a digital co-pilot, ensuring that no fatigue-related accident goes unnoticed.

Developed during Hack4Bengal 4.0 – Eastern India's Largest Hackathon, SARATHI addresses one of the deadliest causes of transport-related deaths — fatigue.

---

YouTube Video Link : https://youtu.be/lDf7jZFTieQ

---

Front View	Side View

🚀 Key Features (USPs)

• ✅ Dual-layer Detection: Simultaneously monitors eye behavior, head pose, and body posture
• 🖼 Screenshot Capture: Takes periodic screenshots during alert states
• 📣 Smart Alerts: Audio alarm + visual warnings when fatigue is detected
• 💡 Lightweight: No heavy ML models — optimized for laptops & Raspberry Pi
• 🔋 Low Resource Usage: CPU-efficient, real-time performance
• 🔧 Fully Modular: Easily extendable for dual camera or IoT deployment

---

📊 Impact & Benefits

🚗 Road Safety	🚌 Fleet Monitoring	✈ Pilot Awareness
Reduces fatigue-related accidents	Real-time monitoring for public transport	Can scale to cockpit fatigue detection

---

🧠 Tech Stack

Category	Tools/Technologies
Language	Python
Libraries	OpenCV, MediaPipe, NumPy, Pygame
UI	Streamlit
Audio	Pygame
Utilities	jsonschema, tornado
Future Hardware	Raspberry Pi 4, external buzzer/cam

---

⚙ Installation & Run Guide

🔄 1. Clone the Repository

bash git clone https://github.com/your-username/sarathi.git

bash cd sarathi

🛠 2. Install Requirements

bash pip install -r requirements.txt

▶ 3. Launch the App

bash streamlit run driver_state_detection/app.py

⚠ Make sure your webcam is connected and accessible.

---

🧪 Technical Workflow

SARATHI performs attention analysis using:

• 👁 Eye landmarks to detect prolonged closure or gaze shifts
• 🧠 Head pose (pitch, yaw, roll) to monitor orientation
• 🧍 Posture deviation using shoulder & spine angles

🧠 If any threshold is crossed:

• Screenshot is captured
• Audio alarm is triggered
• Visual status bar updates on Streamlit UI

---

🧗‍♂ Challenges Faced

• ⚠ Real-time processing without GPU
• 🌙 Handling low-light conditions
• ❗ Ensuring accuracy while minimizing false alarms
• 📐 Angle-sensitive posture detection
• 🧠 Calibrating thresholds across multiple face orientations

---

🌱 Future Enhancements

• 📷 Dual camera setup (side + front view) using IoT (e.g., Raspberry Pi)
• 🧘 User posture calibration on startup for custom thresholds
• ✋ Gesture-based alert dismissal using hand detection
• 🚨 SOS Trigger: Auto-message if alert persists beyond limit
• 🌒 Night mode via IR camera support

---

🏁 Developed At

Hack4Bengal 4.0 – June 2025
Team Name: Dot Slash
Team Lead: Dipan Mazumder
Team Member: Subhas Pramanik

---

🕉 Why the Name "SARATHI"?

In ancient Indian tradition, SARATHI means “charioteer” — the one who safely drives the warrior through chaos.
In our context, SARATHI acts as a *modern digital guide, **monitoring attentiveness, **preventing fatigue-induced danger, and *saving lives through intelligence.

---

📬 Contact

📧 Email	💻 GitHub	🔗 LinkedIn
Dipan Mazumder Subhas Pramanik	Dipan Mazumder Subhas Pramanik	Dipan Mazumder Subhas Pramanik

---

Built to protect. Designed to alert. SARATHI is not just code — it's a life-saving companion.