bichuche0705/highway-vehicle-detection-code

Highway Vehicle Detection - Code & Models

A complete vehicle detection system for highway traffic monitoring. This repository contains the trained models, source code, and documentation - ready to use without requiring dataset downloads.

Quick Start

Installation

pip install ultralytics opencv-python numpy

Basic Usage

from ultralytics import YOLO

# Load the trained model
model = YOLO('models/yolov8m_stage2_improved_best.pt')

# Run inference on an image
results = model('path/to/image.jpg')
results[0].show()

# Process video
results = model('path/to/video.mp4', save=True)

Using the Main Application

python main.py

Repository Contents

Trained Models

• models/yolov8m_stage2_improved_best.pt - Final model (recommended)
• models/yolov8m_stage1_smart_best.pt - Stage 1 model (for comparison)

Source Code

• main.py - Complete vehicle detection and counting application
• example_usage.py - Simple usage examples
• requirements.txt - Python dependencies
• test_improved_model.bat - Windows testing script

Fine-tuning Dataset

• finetune_dataset/images/ - 92 fine-tuning images
• finetune_dataset/labels/ - Corresponding annotation files
• finetune_dataset/README.dataset.txt - Dataset information
• finetune_dataset/README.roboflow.txt - Roboflow export info

Configuration

• dataset_configs/main_data.yaml - Main dataset configuration (8 classes)
• dataset_configs/finetune_data.yaml - Fine-tuning dataset configuration

Training Logs

• training_logs/stage2_results.png - Training results visualization
• training_logs/stage2_confusion_matrix.png - Confusion matrix
• training_logs/stage2_results.csv - Detailed training metrics
• training_logs/stage2_val_batch0_pred.jpg - Sample validation predictions

Training Runs Structure

• training_runs/stage1_smart/ - Stage 1 training configuration and weights
  • args.yaml - Training arguments
  • weights/last.pt - Last epoch weights
• training_runs/stage2_improved/ - Stage 2 training configuration and weights
  • args.yaml - Training arguments
  • weights/last.pt - Last epoch weights
  • BoxF1_curve.png - F1 score curve
  • BoxPR_curve.png - Precision-Recall curve
  • labels.jpg - Label distribution visualization

Documentation

• PROJECT_REPORT.md - Complete project documentation
• README.md - This file

Model Performance

Classes Detected

1. auto - Three-wheelers
2. bus - Public transport vehicles
3. car - Passenger cars
4. lcv - Light Commercial Vehicles
5. motorcycle - Two-wheelers
6. multiaxle - Multi-axle heavy vehicles
7. tractor - Agricultural/construction vehicles
8. truck - Heavy vehicles

Training Stages

• Stage 1: Initial training on 8,219 highway images
• Stage 2: Fine-tuning on 92 additional images for improved truck/bus detection

Fine-tuning Dataset Details

• Images: 92 carefully selected highway images
• Focus: Improved detection of trucks and buses
• Classes: Enhanced examples for problematic vehicle types
• Format: YOLO format with bounding box annotations
• Quality: High-quality images with clear vehicle visibility

External Resources

Test Video

Watch the model in action on YouTube: Highway Vehicle Detection Demo

Main Dataset

Download the complete training dataset from Kaggle: Vehicle Detection 8 Classes Dataset

Technical Details

• Architecture: YOLOv8m (Medium)
• Framework: Ultralytics YOLO
• Input: Images/Videos
• Output: Bounding boxes with class labels and confidence scores
• Hardware: CPU/GPU compatible

Usage Examples

Vehicle Detection

from ultralytics import YOLO
import cv2

# Load the final model
model = YOLO('models/yolov8m_stage2_improved_best.pt')

# Detect vehicles in image
results = model('highway_image.jpg')

# Process results
for result in results:
    boxes = result.boxes
    for box in boxes:
        x1, y1, x2, y2 = box.xyxy[0]
        conf = box.conf[0]
        cls = int(box.cls[0])
        class_name = model.names[cls]
        print(f"Detected: {class_name} (confidence: {conf:.2f})")

Video Processing with Counting

# Process video with vehicle counting
results = model('traffic_video.mp4', save=True, save_txt=True)

# The main.py script provides advanced counting and tracking features

Using the Complete Application

# Run the full application with counting and visualization
from main import VehicleCounter

counter = VehicleCounter()
counter.process_video('input_video.mp4', 'output_video.mp4')

Applications

• Highway traffic monitoring
• Vehicle counting and classification
• Traffic flow analysis
• Automated surveillance systems
• Road safety monitoring
• Traffic data collection

Related Repositories

• Full Dataset: highway-vehicle-detection-full - Complete project with datasets and videos
• Model Only: highway-vehicle-detection - Just the trained model

License

MIT License - Free to use for research and commercial purposes

Contributing

This is a research project. For questions or improvements, please contact the author.

Contact

Author: Nguyen Quoc Viet
Repository: https://huggingface.co/bichuche0705/highway-vehicle-detection-code

Citation

If you use this model in your research, please cite:

@misc{highway-vehicle-detection-code,
  title={Highway Vehicle Detection - Code \& Models},
  author={Nguyen Quoc Viet},
  year={2024},
  url={https://huggingface.co/bichuche0705/highway-vehicle-detection-code}
}