Theses - Undergraduate
Permanent URI for this collection
Browse
Recent Submissions
- ItemOpen AccessPersonal Protective Gear Detection Surveillance System(North South University, 2020-12-30) Amzad Hossain Rafi; Md. Adban Akib Protik; Dipto Roy; Dr. Shahnewaz Siddique; 1530520642; 1610523042; 1530639643The world in present days is going through a tough time and every country a suffering from the present pandemic caused by covid-19 and also at present there is no 100% curable vaccine of this disease. Covid-19 is spreading in rapid rate form one person to another. To minimize the spreading of this diseases WHO (world health org) provided various guideline like wearing mask, maintaining social distancing, avoiding any public gathering. But people are not maintaining those guide line mainly in the countries like Bangladesh, India, Pakistan and many other under developed countries where people live from day to day earning. Keeping that entire thing in our mind we tried to develop a surveillance system that monitor if they are taking enough precautions to save themselves from Covid-19. This surveillance system monitor people are having basic personal protective gear like mask, face shield, PPE, gloves not only that this surveillance system will able to detect if any person is violating the guide line and our system is able to take picture of the violator not only that this surveillance system will able the live detection that will help the people who are monitoring the people.
- ItemOpen AccessGame AI Agent using Deep Reinforcement Learning(North South University, 2020-08-30) Md. Rafat Rahman Tushar; Fariha Islam Faiza; Md. Nuruzzaman; Anas Uddin; Dr. Shahnewaz Siddique; 1621450042; 1620518042; 1621557042; 1520143042Nowadays, the field of deep learning is expanding very fast. We can use deep learning strategies in almost all of the decision making, intelligence problems as well as searching from insanely big amounts of data. In this project, we present an AI agent that learns how to play games from visionary input data. We, humans, also play games by seeing the screen where the game is being played. Our games have limited discrete action space. Our goal is to train an AI agent that has no previous idea of the games’ environments but can successfully learn the strategies of playing games and get scores same as humans. Even we have managed to gain scores more than humans. Our model successfully learns how to play games effectively and in a short amount of time. As we are working with image data, we have to consider the difficulties of processing and training that huge amount of data. Eventually, by implementing a model based on some deep learning strategies, we were able to reach our goal.
- ItemOpen AccessTowards the Analysis and Detection of MS and PhD Admission of Bangladeshi Students into different Ranking University(North South University, 2021-08-30) Md.Fahad Arafin; Md. Faysal Ahmed; Porinita Haque; Dr. Mahdy Rahman Chowdhury; 1520319042; 1521094642; 1711204042Many Bangladeshi students intend to pursue higher studies abroad after completing their undergraduate degrees every year. Choosing a university for higher education is an ambiguous task for students. Usually, they face various problems in selecting the perfect university for them according to their profile. Especially, the students with average and lower academic credentials (undergraduate grades, English proficiency test scores, job, and research experiences) can hardly choose the universities that could match their profile. In this paper, we have analyzed some real unique data of Bangladeshi students who had been accepted admissions at different universities worldwide for higher studies. Finally, we have produced prediction models, which can predict appropriate universities of specific classes for students according to their past academic performances. Two separate models have been studied in this paper, one for MS students and another for PhD students. According to the QS World University Rankings, the universities where the students got admitted have been divided into nine classes for Masters (MS) students and eight classes for PhD students. Random Forest and Decision tree algorithms are used for making the multi-class classification models. F1-score, accuracy, weighted accuracy, and the receiver operating characteristic curves have been studied for the two machine learning algorithms. Numerical results show that for MS data using random forest and decision tree we got same accuracy which is 86%. Again for PhD data using random forest and decision tree we got same accuracy which is 89%.
- ItemOpen AccessCable Operator Management System(North South University, 2020-04-30) Rafid Affan; Shah Md. Munemu Islam; Md. Redwan Muntasir; DR. MOHAMMAD MONIRUJJAMAN KHAN; 1511251642; 1520889042; 1430796042Having a Website will be more convenient for customers and leads and also a website can build better relationships with customers. The 'Cable Providers' are getting more demanding day by day on the market side. Yet our owners are selling an outdated system to clients that are even more annoying. The 'Cable Operator' service provider, let alone the smart payment system, is almost non-digital. Customers still could not notice the Cable Provider material when they wanted it. There is still no proper method and approach for data like that. By providing a web-based management system, this site can easily provide these solutions to the owners and the customers. As well different companies begin their Cable operators business and become an important part of economics, also increasing competition through this field. For this cause of online dependence, this website is built for cable operators to increase their size. In our region, cable operator operation is becoming a wide enterprise. Even in deep rural areas, there is already a link to the cable operator through the televisions that have multicultural activities. Mainly this website is created along with web portal by our project properly. The goal is to ensure a proper management system creation for the owner of this field and also enable them to satisfying their customers as well. This „Cable Operator Management System‟ will create a new era in the field of cable operation.
- ItemOpen AccessA Machine Learning Based Smart Android Application for Visually, Speech, and Hearing Impaired People(North South University, 2020-08-30) Sadman Hossain Ridoy; Tanmim Shikder; AKM BAHALUL HAQUE; 1610456042; 1610694642A Machine Learning based smart application which will be helped for impaired people such as Visual Impaired, Speech Impaired, Hearing Impaired people in their day to day life. Whenever any object or person will be in front of the impaired people or the app will instantly notify the user. Also hearing impaired people also get the alert when someone make any sound to him/her. Speech impaired people can say anything by using this machine trained smart app.