HEAD

💤 Deep Drowsiness Detection using YOLOv5

This project implements a real-time drowsiness detection system using YOLOv5 and webcam video. The model detects faces and classifies them as awake or drowsy, with bounding boxes and confidence scores drawn live on-screen.

---

📂 Project Structure

deep_drowsiness_yolov5/
├── data/
│   ├── images/       # Collected training images
│   └── labels/       # YOLOv5 annotation labels (txt)
├── yolov5/           # YOLOv5 cloned repo (optional if training locally)
├── labelImg/         # Annotation tool (if used)
├── main.py           # Main project code: collection, training, inference
├── yolov5s.pt        # Pretrained YOLOv5 small model (optional)
└── README.md         # You're reading this!

---

📸 Sample Results

Awake (confidence: 0.9):

Mixed Awake & Drowsy Samples (live webcam detections):

Drowsy (bounding box only):

Drowsy with confidence (0.9):

---

🔧 Features

• ✅ Real-time face detection using webcam
• ✅ Custom YOLOv5 model trained to detect "awake" and "drowsy" states
• ✅ Image collection script from webcam
• ✅ Fully annotated and trained using YOLOv5 format
• ✅ Live annotated bounding boxes with class labels and confidence

---

🚀 How to Run

Make sure you have Python and PyTorch installed

1. Clone the repository

git clone https://github.com/yourusername/deep_drowsiness_yolov5.git
cd deep_drowsiness_yolov5

2. Create virtual environment (optional but recommended)

python -m venv venv
source venv/bin/activate  # or venv\Scripts\activate on Windows

3. Install dependencies

pip install -r requirements.txt

4. Collect Training Data

python main.py  # starts data collection (label with keys)

5. Train YOLOv5 Model

cd yolov5
python train.py --img 640 --batch 16 --epochs 30 --data data.yaml --weights yolov5s.pt

6. Run Real-time Detection

python main.py  # runs live detection

---

🧠 Model Details

• Architecture: YOLOv5s
• Classes: awake, drowsy
• Input size: 640x640
• Framework: PyTorch (Ultralytics YOLOv5)

---

🛠 Tools Used

• YOLOv5
• OpenCV
• LabelImg (annotations)
• Python + Matplotlib

---

📜 License

This project is for educational and research use only. (Added README, .gitignore, requirements and sample results)