YOLOv5 Road Safety Helmet Detection

A computer vision project for detecting motorcycle riders, helmets, and number plates using YOLOv5. This system can identify riders wearing helmets, those without helmets, and capture number plates for road safety enforcement.

🎯 Project Overview

This project uses YOLOv5 to detect and classify: - With Helmet: Riders wearing protective helmets - Without Helmet: Riders not wearing helmets - Rider: General rider detection - Number Plate: Vehicle license plate detection

📋 Prerequisites

• Python 3.8 or higher
• CUDA-compatible GPU (recommended for training)
• Git
• WSL2 (if using Windows)

🚀 Installation

1. Clone YOLOv5 Repository

bash git clone https://github.com/ultralytics/yolov5.git cd yolov5

2. Create Virtual Environment (Recommended)

```bash

Using conda

conda create -n yolov5 python=3.8 conda activate yolov5

Or using venv

python -m venv yolov5-env source yolov5-env/bin/activate # On Windows: yolov5-env\Scripts\activate ```

3. Install Dependencies

bash pip install -r requirements.txt

4. Verify Installation

bash python detect.py --source data/images --weights yolov5s.pt --conf 0.25

📁 Dataset Structure

Organize your dataset inside the yolov5 directory:

yolov5/ ├── data/ │ ├── train/ │ │ ├── images/ │ │ └── labels/ │ ├── val/ │ │ ├── images/ │ │ └── labels/ │ └── helmet_data.yaml ├── detect.py ├── train.py ├── val.py └── yolov5s.pt

Dataset Configuration (data/helmet_data.yaml)

Create your dataset YAML file inside the data directory:

```yaml

data/helmet_data.yaml

train: data/train/images val: data/val/images

nc: 4 names: ["with helmet", "without helmet", "rider", "number plate"] ```

🏋️ Training

Basic Training Command

```bash

Navigate to yolov5 directory first

cd yolov5

Train with your helmet dataset

python train.py --img 640 --batch 16 --epochs 50 --data data/helmetdata.yaml --weights yolov5s.pt --name helmetmodel --hyp data/hyps/hyp.scratch.yaml ```

Training Parameters Explained

• --img 640: Input image size
• --batch 16: Batch size (adjust based on GPU memory)
• --epochs 50: Number of training epochs
• --data data/helmet_data.yaml: Path to your custom dataset YAML file
• --weights yolov5s.pt: Pre-trained weights
• --name helmet_model: Experiment name
• --hyp data/hyps/hyp.scratch.yaml: Hyperparameters file

Advanced Training Options

```bash

For better performance with more epochs

python train.py --img 640 --batch 16 --epochs 100 --data data/helmetdata.yaml --weights yolov5s.pt --name helmetmodel_v2 --hyp data/hyps/hyp.scratch.yaml --patience 10

For transfer learning with different model sizes

python train.py --img 640 --batch 8 --epochs 50 --data data/helmetdata.yaml --weights yolov5m.pt --name helmetmodel_medium

Resume training from checkpoint

python train.py --resume runs/train/helmet_model/weights/last.pt ```

📊 Data Preparation

Setting Up Your Dataset

1. Create the directory structure: bash cd yolov5 mkdir -p data/train/images data/train/labels data/val/images data/val/labels

2. Copy your images and labels: ```bash

   Copy training images

   cp /path/to/your/train/images/* data/train/images/

Copy training labels

cp /path/to/your/train/labels/* data/train/labels/

Copy validation images

cp /path/to/your/val/images/* data/val/images/

Copy validation labels

cp /path/to/your/val/labels/* data/val/labels/ ```

1. Verify your dataset: bash # Check if images and labels match python -c " import os train_imgs = len(os.listdir('data/train/images')) train_lbls = len(os.listdir('data/train/labels')) val_imgs = len(os.listdir('data/val/images')) val_lbls = len(os.listdir('data/val/labels')) print(f'Train: {train_imgs} images, {train_lbls} labels') print(f'Val: {val_imgs} images, {val_lbls} labels') "

🔍 Inference

Detect on Images

```bash

From yolov5 directory

python detect.py --source path/to/images --weights runs/train/helmet_model/weights/best.pt --conf 0.4 --save-txt --save-conf ```

Detect on Video

bash python detect.py --source path/to/video.mp4 --weights runs/train/helmet_model/weights/best.pt --conf 0.4

Real-time Webcam Detection

bash python detect.py --source 0 --weights runs/train/helmet_model/weights/best.pt --conf 0.4

Test with Sample Data

```bash

Test with COCO128 sample data first

python detect.py --source data/images --weights yolov5s.pt --conf 0.25

Test with your trained model

python detect.py --source data/val/images --weights runs/train/helmet_model/weights/best.pt --conf 0.4 ```

📊 Model Evaluation

Validation

bash python val.py --data data/helmet_data.yaml --weights runs/train/helmet_model/weights/best.pt --img 640

Export Model

```bash

Export to ONNX

python export.py --weights runs/train/helmet_model/weights/best.pt --include onnx

Export to TensorRT

python export.py --weights runs/train/helmet_model/weights/best.pt --include engine --device 0 ```

📈 Monitoring Training

Training results are saved in runs/train/helmet_model/:

• weights/best.pt: Best model weights
• weights/last.pt: Last epoch weights
• results.png: Training metrics plots
• confusion_matrix.png: Confusion matrix
• train_batch*.jpg: Training batch samples

View Training Progress

```bash

Install tensorboard

pip install tensorboard

View training logs

tensorboard --logdir runs/train ```

🛠️ Troubleshooting

Common Issues

1. CUDA Out of Memory ```bash

   Reduce batch size

   python train.py --batch 8 # or smaller ```

2. Dataset Path Issues

   • Ensure paths in YAML file are relative to the YOLOv5 directory
   • Use forward slashes (/) even on Windows

3. Label Format

   • Ensure labels are in YOLO format: class x_center y_center width height
   • Values should be normalized (0-1)

Performance Optimization

```bash

Multi-GPU training

python -m torch.distributed.launch --nprocpernode 2 train.py --batch 32 --data data/helmet_data.yaml --weights yolov5s.pt --device 0,1

Mixed precision training

python train.py --data data/helmet_data.yaml --weights yolov5s.pt --amp ```

📋 Requirements

Key dependencies: - torch>=1.7.0 - torchvision>=0.8.1 - opencv-python>=4.1.1 - Pillow>=7.1.2 - PyYAML>=5.3.1 - requests>=2.23.0 - tqdm>=4.41.0 - tensorboard>=2.4.1 - wandb - seaborn>=0.11.0

🤝 Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request

📄 License

This project is licensed under the GPL-3.0 License - see the LICENSE file for details.

🙏 Acknowledgments

• Ultralytics YOLOv5 for the base framework
• Road safety dataset contributors
• Open source computer vision community

📞 Support

For issues and questions: 1. Check the YOLOv5 documentation 2. Search existing issues on GitHub 3. Create a new issue with detailed information

---

Note: This project is designed for educational and research purposes in road safety. Ensure compliance with local privacy and surveillance regulations when deploying in production environments.