Deep learning based crowd monitoring and person identification system

This thesis is submitted in partial fulfillment of the requirements for the degree of Bachelor of Science in Computer Science and Engineering, 2023.

Dades bibliogràfiques
Autors principals:	Haque, Mohammad Fahim, Sarkar, Dipto, Choudhury, Tawhid Al Muhaimin, Rafi, Samiul Hoque, Rahim, Md Shajidur
Altres autors:	Chakrabarty, Amitabha
Format:	Thesis
Idioma:	English
Publicat:	Brac University 2024
Matèries:	COVID-19 Faster R-CNN YOLOv8 Micro-controller Data mining
Accés en línia:	http://hdl.handle.net/10361/23612

id	10361-23612
record_format	dspace
spelling	10361-236122024-08-21T08:22:34Z Deep learning based crowd monitoring and person identification system Haque, Mohammad Fahim Sarkar, Dipto Choudhury, Tawhid Al Muhaimin Rafi, Samiul Hoque Rahim, Md Shajidur Chakrabarty, Amitabha Department of Computer Science and Engineering, Brac University COVID-19 Faster R-CNN YOLOv8 Micro-controller Data mining COVID-19 This thesis is submitted in partial fulfillment of the requirements for the degree of Bachelor of Science in Computer Science and Engineering, 2023. Cataloged from PDF version of thesis. Includes bibliographical references (pages 51-53). In this paper, we propose a deep learning-based crowd monitoring and person identification system and a crowd-video dataset to address the challenges posed by the recent COVID-19 pandemic or future pandemics may occur, where maintaining social distance in public places is necessary. The system can be also implemented where crowd monitoring is necessary. For instance, public places like bank booths, airports, train stations, hospitals, tourist attractions, public transportation hubs, stadiums, arenas, etc. The system combines person tracking and social distance measurements to accurately detect individuals who may unintentionally violate the rules due to a lack of spatial awareness. To implement the system, a custom dataset was created to evaluate and tackle perspective correction and person-only detection issues. Three popular object detection models: YOLOv8, YOLOv7, and Faster R-CNN with and without the DeepSort tracking algorithm were used and a comparison of their performances is demonstrated. To build our system we took two different approaches. In the first approach, we used Faster R-CNN & YOLOv8 for person identification, and for tracking, we used the SORT tracking algorithm. In the second approach, we used YOLOv7 & YOLOv8 for person detection and DeepSortbased tracking algorithm which generates unique IDs and successfully tracks and reidentifies each person frame by frame using Kalman filter and Hungarian algorithm. The experimental results show that all models can accurately detect humans with 90% accuracy and estimate the distance between them. However, Faster R-CNN falls short in real-time human detection, whereas YOLOv8 outperforms YOLOv7 in terms of speed and detection accuracy. Still, YOLOv8 is relatively new and has less support for implementation. Thus, YOLOv7 is chosen for implementation in mobile, micro-controller-based, or IoT devices, as it offers better support for immediate implementation. The proposed system is efficient, accurate, and does not require human supervision. It includes a log system to track violations with frame rates and unique IDs. The system was tested using our custom dataset, and positive results were achieved, indicating its potential usefulness in crowd monitoring and social distance enforcement. Mohammad Fahim Haque Tawhid Al Muhaimin Choudhury Samiul Hoque Rafi Md Shajidur Rahim Dipto Sarkar B.Sc in Computer Science 2024-06-26T10:51:16Z 2024-06-26T10:51:16Z ©2023 2023-09 Thesis ID 18201141 ID 18201182 ID 18201191 ID 18201178 ID 18101535 http://hdl.handle.net/10361/23612 en Brac University theses are protected by copyright. They may be viewed from this source for any purpose, but reproduction or distribution in any format is prohibited without written permission. 64 pages application/pdf Brac University
institution	Brac University
collection	Institutional Repository
language	English
topic	COVID-19 Faster R-CNN YOLOv8 Micro-controller Data mining COVID-19
spellingShingle	COVID-19 Faster R-CNN YOLOv8 Micro-controller Data mining COVID-19 Haque, Mohammad Fahim Sarkar, Dipto Choudhury, Tawhid Al Muhaimin Rafi, Samiul Hoque Rahim, Md Shajidur Deep learning based crowd monitoring and person identification system
description	This thesis is submitted in partial fulfillment of the requirements for the degree of Bachelor of Science in Computer Science and Engineering, 2023.
author2	Chakrabarty, Amitabha
author_facet	Chakrabarty, Amitabha Haque, Mohammad Fahim Sarkar, Dipto Choudhury, Tawhid Al Muhaimin Rafi, Samiul Hoque Rahim, Md Shajidur
format	Thesis
author	Haque, Mohammad Fahim Sarkar, Dipto Choudhury, Tawhid Al Muhaimin Rafi, Samiul Hoque Rahim, Md Shajidur
author_sort	Haque, Mohammad Fahim
title	Deep learning based crowd monitoring and person identification system
title_short	Deep learning based crowd monitoring and person identification system
title_full	Deep learning based crowd monitoring and person identification system
title_fullStr	Deep learning based crowd monitoring and person identification system
title_full_unstemmed	Deep learning based crowd monitoring and person identification system
title_sort	deep learning based crowd monitoring and person identification system
publisher	Brac University
publishDate	2024
url	http://hdl.handle.net/10361/23612
work_keys_str_mv	AT haquemohammadfahim deeplearningbasedcrowdmonitoringandpersonidentificationsystem AT sarkardipto deeplearningbasedcrowdmonitoringandpersonidentificationsystem AT choudhurytawhidalmuhaimin deeplearningbasedcrowdmonitoringandpersonidentificationsystem AT rafisamiulhoque deeplearningbasedcrowdmonitoringandpersonidentificationsystem AT rahimmdshajidur deeplearningbasedcrowdmonitoringandpersonidentificationsystem
_version_	1814309237364359168

Deep learning based crowd monitoring and person identification system

Ítems similars