This dataset contains annotated images for monitoring helmet usage behavior in traffic conditions, specifically designed for research on "Применение модели глубокого обучения YOLOv11 для мониторинга поведения пользователей шлемов в условиях дорожного движения. Сравнение различных моделей" (Application of Deep Learning Model YOLOv11 for Monitoring Helmet User Behavior in Road Traffic Conditions. Comparison of Different Models)

Uses

Research Applications:

Model Comparison: Performance evaluation of YOLOv8, YOLOv11, and other object detection models
Helmet Usage Analysis: Statistical analysis of helmet compliance in traffic conditions
Traffic Safety Research: Study of correlation between helmet usage and road safety metrics
Deep Learning Benchmark: Benchmark dataset for binary classification in traffic monitoring

Practical Applications:

Traffic law enforcement monitoring systems
Road safety awareness campaigns
Smart city traffic management
Insurance risk assessment

Dataset Creation Methodology

Data Collection:

Source Videos: 13 traffic surveillance videos (source1.mp4 to source13.mp4)
Duration: Various lengths, total approximately [X] hours
Resolution: Original video resolution maintained

Processing Pipeline:

Frame Extraction: 1 frame every 5 seconds using FFmpeg
Auto-labeling: Initial labeling using YOLOv8 pretrained model
Manual Verification: Quality check and correction
Classification: Binary classification (with_helmet/without_helmet)
Dataset Splitting: 70-20-10 split for train-val-test

Quality Assurance:

All images checked for appropriate lighting conditions
Clear distinction between helmet and non-helmet cases
Balanced representation of various traffic scenarios

📈 Performance Metrics (Expected) For model evaluation, consider these metrics:

Accuracy: Overall classification accuracy
Precision: Proportion of correctly identified helmets
Recall: Ability to find all helmet instances
Inference Speed: Frames per second (FPS) for real-time applications

License

This dataset is licensed under Creative Commons Attribution 4.0 International (CC-BY-4.0).

You are free to: Share — copy and redistribute the material in any medium or format Adapt — remix, transform, and build upon the material For any purpose, even commercially

Under the following terms: Attribution — You must give appropriate credit, provide a link to the license, and indicate if changes were made Full license text: https://creativecommons.org/licenses/by/4.0/

Limitations & Bias

Known Limitations:

Geographic Bias: Data collected primarily from Vietnamese traffic conditions
Weather Conditions: Limited representation of extreme weather
Time of Day: May not equally represent all lighting conditions
Vehicle Types: Focus on motorcycles; limited other vehicle types

Ethical Considerations:

All personal identifiers removed from images
Used for research purposes only
Compliant with privacy regulations
Non-intrusive surveillance methodology

Resolution: Original video resolution maintained

Processing Pipeline:

Frame Extraction: 1 frame every 5 seconds using FFmpeg
Auto-labeling: Initial labeling using YOLOv8 pretrained model
Manual Verification: Quality check and correction
Classification: Binary classification (with_helmet/without_helmet)
Dataset Splitting: 70-20-10 split for train-val-test

Quality Assurance:

All images checked for appropriate lighting conditions
Clear distinction between helmet and non-helmet cases
Balanced representation of various traffic scenarios

📈 Performance Metrics (Expected) For model evaluation, consider these metrics:

Accuracy: Overall classification accuracy
Precision: Proportion of correctly identified helmets
Recall: Ability to find all helmet instances

Inference Speed: Frames per second (FPS) for real-time applications

License

This dataset is licensed under Creative Commons Attribution 4.0 International (CC-BY-4.0).

You are free to:

Share — copy and redistribute the material in any medium or format Adapt — remix, transform, and build upon the material For any purpose, even commercially

Under the following terms:

Attribution — You must give appropriate credit, provide a link to the license, and indicate if changes were made Full license text: https://creativecommons.org/licenses/by/4.0/

Limitations & Bias

Known Limitations: Geographic Bias: Data collected primarily from Vietnamese traffic conditions Weather Conditions: Limited representation of extreme weather Time of Day: May not equally represent all lighting conditions Vehicle Types: Focus on motorcycles; limited other vehicle types

Ethical Considerations: All personal identifiers removed from images

Used for research purposes only

Compliant with privacy regulations

Non-intrusive surveillance methodology

Contact

For questions, collaboration, or dataset-related inquiries:

Corresponding Author: [Pham Minh Hieu, Le Nhu Ngoc, Giang Khac Ngu]

Email: [m2011254@edu.misis.ru m1911859@edu.misis.ru m2011254@edu.misis.ru]

Institution: [The National University of Science and Technology MISIS]

Research Group: [МПИ-25-1-2]

Downloads last month: 116

Total file size:

974 MB