https://github.com/dipan313/sarathi
🚘 SARATHI – A real-time AI-based attention monitoring system for drivers and pilots that detects drowsiness, posture issues, and inattention. Built with Python, MediaPipe, OpenCV, and Streamlit.
https://github.com/dipan313/sarathi
blinker matplotlib mediapipe opencv pygame python raspberry-pi streamlit t
Last synced: 2 months ago
JSON representation
🚘 SARATHI – A real-time AI-based attention monitoring system for drivers and pilots that detects drowsiness, posture issues, and inattention. Built with Python, MediaPipe, OpenCV, and Streamlit.
- Host: GitHub
- URL: https://github.com/dipan313/sarathi
- Owner: dipan313
- Created: 2025-06-21T07:56:20.000Z (12 months ago)
- Default Branch: main
- Last Pushed: 2025-06-21T18:35:35.000Z (12 months ago)
- Last Synced: 2025-06-21T19:25:05.426Z (12 months ago)
- Topics: blinker, matplotlib, mediapipe, opencv, pygame, python, raspberry-pi, streamlit, t
- Language: Python
- Homepage:
- Size: 79.1 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# 🚘 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](mailto:dipanmazumder313@gmail.com)
[Subhas Pramanik ](mailto:subhaspramanik38@gmail.com) | [Dipan Mazumder](https://github.com/dipan313)
[Subhas Pramanik](https://github.com/subhas-pramanik-09) | [Dipan Mazumder](https://www.linkedin.com/in/dipan-mazumder-953453279/)
[Subhas Pramanik](https://www.linkedin.com/in/subhas-pramanik) |
---
> Built to protect. Designed to alert. SARATHI is not just code — it's a life-saving companion.