Synopsis

Helmet recognition using Machine learning approach

 HELMET RECOGNITION USING MACHINE LEARNING
APPROACH
A MINOR PROJECT SUBMITTED
to

CHHATTISGARH SWAMI VIVEKANAND TECHNICAL

UNIVERSITY
BHILAI (C.G.), INDIA
In Partial Fillment
For the award ofree
Of
BACHELOR IN ENGINEERING
In
Computer Science and Engineering
by

Gaurav Sahu(CB2702, 313302221009)

Himangee Tripathi(CB2716, 313302221020)
Sonal Sahu(CB2721, 313302221023)
Dhaneshwari Painkra(CB2736, 313302221031)
Vandana Kaushik(CB2714, 313302221055)

Under the Guidance of

Dr. Mohan Rao Mamdikar
Assistant Professor
Department of Computer Science and Engineering
Vishwavidyalaya Engineering College Ambikapur
(A constituent College of Chhattisgarh Swami Vivekanand Technical University, Bhilai)
___________________________________________________
Department of Computer Science and Engineering
Vishwavidyalaya Engineering College Ambikapur
(A constituent College of Chhattisgarh Swami Vivekanand Technical University, Bhilai)
___________________________________________________
Session: 2024 – 2025
 DECLARATION BY THE CANDIDATE

We undersigned solemnly declare that the report of the thesis work

entitled HELMET RECOGNITION USING MACHINE LEARNING APPROACH is
based our own work carried out during the course of our study under the
supervision of Dr. Mohan Rao Mamdikar.

We assert that the statements made and conclusions drawn are an

outcome of the project work. We further declare that to the best of our
knowledge and belief that the report does not contain any part of any work
which has been submitted for the award of any other degree or certificate in this
University of India or any other country. All helps received and citations used
for the preparation of thesis have been duly acknowledged.

Gaurav Sahu(313302221023)
Himangee Tripathi(313302221020)
Sonal Sahu(313302221023)
Dhaneshwari Painkra(313302221031)
Vandana Kaushik(313302221055)
Dr. Mohan Rao Mamdikar
Assistant Professor
Department of Computer Science and Engineering
Vishwavidyalaya Engineering College, Ambikapur
 CERTIFICATE BY THE SUPERVISOR

This is to certify that the report of the thesis entitled “HELMET RECOGNITION USING
MACHINE LEARNING APPROACH” is a record of bona-fide research work carried out by
Gaurav Sahu, Himangee Tripathi, Sonal Sahu, Dhaneshwari Painkra, Vandana Kaushik,
bearing Roll No.: 313302221009, 313302221020, 313302221023, 313302221031,
313302221055, & Enrollment No.: CB2702, CB2716 CB2721, CB2736, CB2714,
respectively under my guidance and supervision for the award of Bachelor of
Engineering, in Computer Science and Engineering of Chhattisgarh Swami Vivekanand
Technical University, Bhilai (C.G.) India.

To the best of my Knowledge the thesis

 Embodies the work of the candidate him/herself,
 Has duly been completed,
 Fulfill the requirements of the ordinance relating to the M.E. degree of the
University and
 Is up to the desired standard both in respect of contents and language for
being referred to the examiners
Dr. Mohan Rao Mamdikar
Assistant Professor
Department of Computer Science and Engineering
Vishwavidyalaya Engineering College, Ambikapur
 DISSERTATION APPROVAL SHEET

Report entitled “HELMET RECOGNITION USING MACHINE LEARNING APPROACH” by

Gaurav Sahu, Himangee Tripathi, Sonal Sahu, Dhaneshwari Painkra, Vandana
Kaushik, bearing Roll No.: 313302221009, 313302221020, 313302221023, 313302221031,
313302221055, & Enrollment No.: CB2702, CB2716 CB2721, CB2736, CB2714,
respectively for the fulfillment of Bachelor of Engineering, in Computer Science and
Engineering is approved & forwarded to Chhattisgarh Swami Vivekanand Technical
University, Bhilai (C.G.), India.

Forwarded to Chhattisgarh Swami Vivekanand Technical University, Bhilai C. G.

(Project Supervisor)
Dr. Mohan Rao Mamdikar
Assistant Professor
Department of Computer Science and Engineering
Vishwavidyalaya Engineering College, Ambikapur
 (Head of Department)
Ms. Jasmine Minj
Assistant Professor
Department of Computer Science and Engineering
Vishwavidyalaya Engineering College, Ambikapur

(PRINCIPAL)
Dr. R.N. Khare
Vishwavidyalaya Engineering College Ambikapur
(A constituent College of Chhattisgarh Swami Vivekanand Technical University, Bhilai)
 ACKNOWLEDGEMENT

We want to show our genuine thanks to everyone who helped make our project, "Helmet Detection Using
Machine Learning," a success.

Foremost, we owe a great deal of gratitude to our project mentor, Dr. Mohan Rao Mamdikar for his
priceless assistance, backing, and motivation during this project. His knowledge and understanding have
played a crucial role in forming this project.

We are thankful to the Department of Computer Science and Engineering at Vishwavidyalaya Engineering
College, Ambikapur for supporting us with the essential resources and facilities for conducting this research.
The department's dedication to excellence and innovation has fostered a suitable setting for learning and
progress.

We want to express our gratitude to all the department's faculty and staff for their assistance and motivation.
Their commitment to both teaching and research has motivated us to strive towards our academic objectives.

Finally, we would like to acknowledge the contributions of the open-source community, whose libraries and
frameworks have been essential to the development of this project. In particular, we am grateful for the work
done on OpenCV, Python, Image Processing, HTML, CSS, MySQL, Matplotlib, and Tkinter. These tools
have provided us with the necessary capabilities to implement and evaluate our helmet detection model.

Gaurav Sahu(313302221023)
Himangee Tripathi(313302221020)
Sonal Sahu(313302221023)
Dhaneshwari Painkra(313302221031)
Vandana Kaushik(313302221055)
 Abstract
This documentation outlines a minor project focused on helmet detection and recognition using advanced
computer vision techniques. The project employs the YOLOv8 (You Only Look Once) object detection
algorithm to identify helmets in images and videos, aiming to enhance safety compliance in environments
such as construction sites and road traffic. The proposed system processes input images or video streams,
detects the presence of helmets, and generates annotated outputs alongside performance metrics. This project
serves as a foundational step toward developing automated safety monitoring systems that can significantly
reduce accidents and injuries related to helmet non-compliance.
Table of Contents
1. Abstract
2. List of Abbreviations
3. List of Figures
4. List of Tables
5. Chapter 1: Introduction
5.1 Overview
5.2 Objectives
5.3 Problem Statement
6. Chapter 2: Literature Review
6.1 Related Work
6.2 Existing Technologies
7. Chapter 3: Methodology
7.1 System Architecture
7.2 Tools and Technologies
7.3 Implementation Steps
8. Chapter 4: Results and Discussion
8.1 Performance Metrics
8.2 Evaluation
9. Chapter 5: Conclusion and Future Work
10. References
11. List of Abbreviations
YOLO: You Only Look Once
CNN: Convolutional Neural Network
SSD: Single Shot Multibox Detector
CV: Computer Vision
API: Application Programming Interface
12. List of Figures
Figure 1: System Architecture Diagram
Figure 2: Example of Detected Helmets
Figure 3: Confusion Matrix of Detection Results
13. List of Tables
Table 1: Performance Metrics of YOLOv8 Model
Table 2: Comparison of Detection Algorithms
Table 3: Dataset Summary
Chapter 1: Introduction
1.1 Overview
The increasing number of accidents involving motorcyclists and construction workers not wearing helmets
highlights a critical safety concern. This project aims to address this issue through the development of a
helmet detection system that utilizes state-of-the-art object detection algorithms. By automating the detection
of helmet usage, this system can provide real-time monitoring and enhance compliance with safety
regulations.
1.2 Objectives
The primary objectives of this project include:
 To develop a robust helmet detection system using the YOLOv8 algorithm.
 To evaluate the performance of the detection system in various scenarios, including different lighting
and environmental conditions.
 To create a user-friendly interface for displaying detection results and performance metrics.
1.3 Problem Statement
Despite the availability of safety equipment, many individuals fail to wear helmets, leading to preventable
injuries and fatalities. Current monitoring methods are often manual and inefficient. This project seeks to
create an automated solution that can accurately detect helmet usage in real-time, thereby promoting safer
practices in high-risk environments.
Chapter 2: Literature Review
2.1 Related Work
Numerous studies have explored helmet detection using various machine learning and deep learning
techniques. Previous works have demonstrated the effectiveness of algorithms such as SSD[1] and Faster R-
CNN in detecting safety equipment. However, there remains a gap in real-time applications tailored for
specific environments like construction sites and traffic monitoring .
2.2 Existing Technologies
Technologies such as OpenCV and deep learning frameworks like TensorFlow and PyTorch have been widely
used in image processing and object detection tasks. The YOLOv8 model, in particular, has shown promise
due to its speed and accuracy, making it suitable for real-time applications.
For object detection, the collection of images containing each object is used to train the custom model. Once
the model is formed, it can be used to detect these five objects. We need to collect a large data set of cyclists
with and without helmets.[2] After the collection, we need to mark each image as needed. When the
annotation is complete, the images will be fed into the Yolo v3 model, which is used to form custom classes.
The model is loaded with the weights obtained after training. Next, an image or video input is provided. In
this case, the five classes formed will be detected by the system. We can quickly find information about the
primary class of the cyclist if the person does not wear a helmet.[1]
Chapter 3: Methodology
3.1 System Architecture
The system architecture consists of several components, including:
 Input Module: Captures images or video streams.
 Processing Module: Implements the YOLOv8 algorithm for helmet detection.
 Output Module: Displays results and generates reports.

3.2 Tools and Technologies

Programming Language: Python
Libraries: OpenCV, YOLOv8, Supervision, Ultralytics
3.3 Implementation Steps
 Load the pre-trained YOLOv8 model.
 Capture input images or video frames.
 Process frames to detect helmets.
 Visualize detection results and save annotated outputs.
 Evaluate performance using metrics such as accuracy and confusion matrix.
Chapter 4: Results and Discussion
4.1 Performance Metrics
The performance of the helmet detection system is evaluated using metrics such as precision, recall, and F1-
score. The results indicate that the YOLOv8 model achieves high accuracy in detecting helmets under various
conditions.
4.2 Evaluation
The evaluation includes a comparison of detection results against ground truth data, highlighting the model's
strengths and areas for improvement. The confusion matrix provides insights into the model's performance,
including false positives and negatives.
Chapter 5: Conclusion and Future Work
This project successfully demonstrates the feasibility of using YOLOv8 for helmet detection in real-time
applications. Future work may involve expanding the system to detect additional safety equipment and
improving accuracy through model fine-tuning with larger datasets.
References
[1] Pathasu Doungmala, Katanyoo Klubsuwan, “Half and Full Helmet Wearing Detection in Thailand using
Haar Like Feature and Circle Hough Transform on Image Processing”, IEEE International Conference on
Computer and Information Technology.IEEE, 2016.
[2] Jie Li, Huanming Liu, Tianzheng Wang ,Min Jiang and Kang Li, “Safety Helmet Wearing Detection
Based on Image Processing and Machine Learning”, Ninth International Conference on Advanced
Computational Intelligence(ICACI), pp.109-119, 2017.