🚗 Real-Time Driver Drowsiness Detection System

📌 Problem Statement

Driver fatigue and drowsiness are among the leading causes of road accidents, especially during night-time long-haul driving such as truck and highway transport. Traditional safety measures fail because they do not monitor the driver’s physical alertness in real time.

There is a strong need for a real-time, automated, and intelligent system that can:

• Continuously monitor a driver
• Detect whether the driver is awake or drowsy
• Trigger alerts before an accident occurs
• Be extendable to IoT-based safety systems

---

💡 Our Solution

This project implements a real-time drowsiness detection system using Deep Learning and Computer Vision. Working Video

Key Idea

• Train a binary image classification model (Awake vs Drowsy)
• Use a pretrained ResNet-18 for robust feature extraction
• Perform live face detection via webcam
• Predict driver state in real time
• Raise alerts if drowsiness persists

Why this works

• CNNs capture facial fatigue patterns
• Transfer learning improves accuracy with limited data
• Temporal frame tracking reduces false positives

---

🧠 System Architecture

Pipeline Flow:

Webcam Frame
     ↓
Face Detection (OpenCV Haar Cascade)
     ↓
Face Cropping & Preprocessing
     ↓
ResNet-18 Classifier
     ↓
Awake / Drowsy Prediction
     ↓
Alert Trigger (if drowsy persists)

---

🛠️ Tech Stack

• Language: Python
• Deep Learning: PyTorch, Torchvision
• Model: ResNet-18 (Transfer Learning)
• Computer Vision: OpenCV
• Image Processing: Pillow, NumPy
• Dataset Handling: YOLO-style labels
• Deployment Ready: IoT & Edge-compatible

---

📂 Project Structure

├── datapreprocessing.py        # Dataset loading & preprocessing
├── ResNetmodel.py              # Model training & saving
├── runmodel.py                 # Real-time webcam inference
├── requirements.txt            # Project dependencies
├── drowsiness_detection_model.pth  # Trained model (generated)
└── README.md

---

⚙️ How to Run the Project

1️⃣ Install Dependencies

pip install -r requirements.txt

---

2️⃣ Dataset Preparation

Ensure the dataset follows this structure:

images/
 ├── img1.jpg
 ├── img2.jpg

labels/
 ├── img1.txt
 ├── img2.txt

Label format (YOLO-style):

0 → Awake
1 → Drowsy

---

3️⃣ (Optional) Preprocess Data

python datapreprocessing.py

This step:

• Splits dataset into train/test
• Saves metadata and NumPy arrays (optional)

---

4️⃣ Train the Model

python ResNetmodel.py

This will:

• Train ResNet-18 on your dataset
• Save the trained model as:

drowsiness_detection_model.pth

---

5️⃣ Run Real-Time Detection

python runmodel.py

Controls:

• Webcam opens automatically
• Press q to exit

---

🚨 Output Behavior

• Green Box: AWAKE
• Red Box: DROWSY
• Alert Trigger: If drowsiness continues for multiple frames, a DROWSINESS ALERT is displayed.

---

🌐 Real-World Applications

• 🚛 Truck & Bus Driver Monitoring

• 🚗 Night-time highway safety

• 🏭 Industrial machine operator alertness

• 🧠 Workplace fatigue monitoring

• 🔌 IoT integration with:

  • Buzzers
  • Seat vibration
  • Smart dashboards
  • Fleet monitoring systems

---

🔮 Future Enhancements

• Eye Aspect Ratio (EAR) + blink analysis
• Temporal models (LSTM / sliding window)
• Face detection upgrade (MTCNN / RetinaFace)
• Edge deployment (Jetson / Raspberry Pi)
• IoT alerts (GSM / MQTT / CAN bus)
• Driver fatigue analytics dashboard

---