Enhancing School Discipline with AI-Based Dress Code Monitoring System
Author: SivaGuru.K
Institution: R P Sarathy institute of technology, Salem.
Email: [email protected]
Abstract
School dress codes play a crucial role in maintaining discipline, uniformity, and
professionalism among students. However, manual enforcement of dress codes is often
inconsistent and time-consuming. This paper presents an AI-based dress code monitoring
system that utilizes computer vision and deep learning to automatically detect dress code
violations, such as untucked shirts, missing ties, improper shoes, and unkempt hairstyles.
The system is integrated with student ID scanners and a MySQL database to track
compliance and notify students in real time. The study evaluates different machine learning
models and discusses ethical concerns related to AI-based surveillance. The results
demonstrate that AI can significantly enhance school discipline by providing a fair,
automated, and scalable solution.
1. Introduction
1.1 Background & Motivation
Maintaining a proper dress code in schools is essential for ensuring discipline, equality, and
a professional learning environment. Traditional enforcement methods rely on manual
inspections by teachers and staff, which often lead to inconsistencies, subjectivity, and
inefficiencies. Some key challenges include:
- Time-consuming checks: Manually verifying each student’s attire takes significant effort.
- Inconsistent enforcement: Different staff members may interpret dress code rules
differently.
- Difficulty in tracking violations: Repeated offenders may go unnoticed without proper
record-keeping.
1.2 Objective
This study aims to develop an AI-based dress code monitoring system that can:
1. Detect dress code violations (untucked shirts, missing ties, improper shoes, unkempt
hairstyles) using deep learning.
2. Integrate with student ID scanners to track compliance.
3. Log compliance records in a MySQL database for analysis.
4. Notify students and school authorities in real time.
2. Literature Review
2.1 AI and Computer Vision in Education
Recent research highlights the growing use of AI-driven monitoring systems in schools,
including automated attendance, behavior tracking, and security surveillance. Studies show
that deep learning models like YOLO, Faster R-CNN, and ResNet are highly effective for
object detection tasks, making them suitable for dress code enforcement.
2.2 Existing Dress Code Monitoring Methods
- Manual inspections: Time-consuming and prone to human error.
- RFID-based systems: Track uniforms but do not analyze compliance visually.
- AI-powered surveillance in workplaces: Successfully used for monitoring employee attire
but not widely adopted in schools.
2.3 Ethical and Privacy Considerations
AI-based monitoring must respect student privacy and be transparent in its
implementation. Schools must implement secure data handling practices to prevent misuse.
3. Methodology
3.1 System Architecture
The proposed system consists of:
1. Camera Modules – Capture images of students at school entry points.
2. AI-Based Dress Code Detection – Uses YOLOv5 for real-time analysis.
3. Student ID Integration – Links violations to individual student records.
4. Notification System – Sends alerts via SMS or school apps.
5. Database Logging – Stores data in MySQL for record-keeping.
3.2 Data Collection & Preprocessing
Dataset Creation: Images of students in both compliant and non-compliant attire were
collected under various lighting conditions and angles.
Annotation: Labeled images for different dress code violations using LabelImg.
Data Augmentation: Applied rotation, brightness adjustment, and noise addition to improve
model robustness.
3.3 AI Model Selection & Training
Compared YOLOv5, Faster R-CNN, and ResNet for dress code detection.
YOLOv5 was selected due to its high accuracy (94%) and real-time processing speed (30
fps).
The model was trained on a custom dataset using transfer learning.
3.4 System Integration
AI model was deployed on a cloud server with API-based access.
Linked to a MySQL database storing student records and violations.
Developed a web dashboard for school administrators to monitor compliance.
4. Results & Discussion
4.1 Model Performance
Key Observations:
- YOLOv5 demonstrated the highest performance, making it the best choice for real-time
monitoring.
- The model faced false positives in extreme lighting conditions, which can be improved
through additional training.
4.2 Ethical Considerations
Student Privacy: Ensured secure data storage with restricted access.
AI Fairness: The model was trained on a diverse dataset to avoid biases.
Transparency: Schools must inform students and parents about the system’s
implementation.
5. Conclusion & Future Work
This study demonstrates that an AI-based dress code monitoring system can significantly
enhance school discipline by:
- Reducing manual enforcement workload.
- Ensuring fair and unbiased compliance checks.
- Providing real-time alerts for immediate correction.
Future Enhancements:
- Expanding dataset with more diverse student appearances.
- Integration with CCTV systems for broader monitoring.
- Adding voice alerts for real-time compliance guidance.
6. References
1. Redmon, J., & Farhadi, A. (2018). 'YOLOv3: An Incremental Improvement.' arXiv.
2. He, K., Zhang, X., Ren, S., & Sun, J. (2016). 'Deep Residual Learning for Image Recognition.'
IEEE CVPR.
3. Goodfellow, I., Bengio, Y., & Courville, A. (2016). 'Deep Learning.' MIT Press.
4. Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). 'ImageNet Classification with Deep
Convolutional Neural Networks.' Advances in Neural Information Processing Systems
(NeurIPS).
5. Simonyan, K., & Zisserman, A. (2015). 'Very Deep Convolutional Networks for Large-Scale
Image Recognition.' International Conference on Learning Representations (ICLR).
6. Huang, G., Liu, Z., Van Der Maaten, L., & Weinberger, K. Q. (2017). 'Densely Connected
Convolutional Networks.' IEEE CVPR.
7. Howard, A. G., et al. (2017). 'MobileNets: Efficient Convolutional Neural Networks for
Mobile Vision Applications.' arXiv.
8. Dosovitskiy, A., et al. (2020). 'An Image is Worth 16x16 Words: Transformers for Image
Recognition at Scale.' arXiv.
