İremnur Kulaksız
Automatic Annotation and Labeling of Fashion Images
Automatic Annotation and Labeling of Fashion Images is a research project. Visual fashion analysis is becoming more and more important in recent years for the fashion industry because of its huge potential. Main motivation of this project is predicting clothing category using VGG16 and RESNET50. In this project, as the number of classes changes, the success rates of these networks were analyzed on 2 different data sets.

Tunahan Yadigarbigün
Statistical Analysis of Fingerprint Images
Fingerprints are being used in several areas on the purpose of identity recognition thanks to their individuality and permanence features.  In this project, it is aimed to increase the success in fingerprint recognition processes by performing statistical analysis in order to identify the low quality data in those large databases and at the same time to benefit from the distinctiveness of the high quality data.

Umut Utku Tahan
Scene Text Recognition
Scene text recognition leads to many applications like visual question answering, automated receipt reading, indexing of multimedia archives, identifying vehicles by reading their license plates etc. Scene text recognition is getting a lot of attention by researchers and there are competitions for improving the success rate of it.  Our goal in this research is to improve the success rate of detecting scene text.

Ali Acar, Bora Baloğlu
Terms-Definitions Extractor from a Text
There are 2 phases in our thesis. 1: Definitional Sentence Analysis which we tried to decide whether a sentence has definition or not with textbook data. 2: StackExchange Answerer Analysis which we tried to create quality assessment tool for Grey Literature of Software Engineering with StackExchange users’ answers. In both phases we use natural language processing methods with supervised learning.

Elgun Jabrayilzade
Solving Logic Grid Puzzles Automatically
Logic grid puzzle is a type of word problem where the task is to solve a problem in logic. Constraints for the problem are given in the form of textual clues. LGPSolver is a system that automatically solves logic grid puzzles. Better identification of comparison structures in clues and the use of a DistilBERT-based clue classification solution are the two highlights of the system. LGPSolver achieves full accuracy in sample datasets.

Cem Sakızcı
Efficient Implementation of Graph Coloring on GPU
The aim of this project is to implement an efficient algorithm, named as Counting-based Jones-Plassmann algorithm (CJP), for solving the graph coloring problem on GPU by utilizing its architecture. We carried out the implementation by using CUDA programming model. By realizing the CJP algorithm, we aimed to achieve high coloring speed and minimum number of colors used compared to existing graph coloring algorithms.

Cemal İhsan Sofuoğlu, Onur Karakoç
Developing a Tracking System for Animals with an Application
A system that will help pet owners to trace or observe their pet from a mobile application. Application will use Global Positioning System (GPS) technology to determine the location. The coordinates will be transfered to the application via Bluetooth Technology. After receiving the coordinates from device, application will parse the latitude and longtitude, then the location will be displayed on the map.

Kıvanç Ersoy, Ertan Uysal, Mücahit Turhan, Ahmet Öcal
LareCog
We see many different landmark on images in both websites and social media and we wonder where these pictures are and what their history is. In this project, we try to solve this problem by developing mobil application which can easily be used by people. We used landmark recognition technique which includes images recognition and images classification with using machine learning techniques.

Dinçer Güner, Egemen Baki
Developing an Application for Customer Analysis
DACA is a project that helps retail store owners to analyze their customers’ feelings about the store. The vision of the project is achieved via a real-time application powered by machine learning. It lists how much time customers spend in stores, how customers’ facial expressions were before they entered the store, and how customers’ facial expressions were after they left the store. The application has also combined detailed analysis through these features.

Bora Özsoy, Busenur Aktılav, Eren Can Güleç, Özgesu Özen
Automated Measurement of Functional Size from Code
Automated Measurement of Functional Size from Code project offers to develop software that measures the functional size of an existing project by using COSMIC function points in order to reduce time and budget costs by predicting the size of future projects. And the Cosmic APP measures the size of the software through sequence diagrams which are produced by using reverse engineering tools.

Onur Sak, Can Çandır, Mehmet Şen
Intelligent Sales Incentive System for E-commerce Sites
In online shopping, the abandonment rate is the ratio of the number of abandoned shopping carts to the number of all transactions which are completed or not. This ratio is become higher over the years and still increasing. Our main objective is to decrease this rate by analyzing customer behavior and encourage the customers to complete their shopping.

Merve Koyuncu, Zekiye Kirişçi
A Simulation of Hybrid Information Spreading Model for Social Networks
This study aims to have more realistic predictions for malware spreading on a social network performing an interaction and behavior analysis among its users. It can also be utilized to personalize the advertisements and to explain how an information spreads on social networks according to categories of the user. It explains how this can be simulated and which algorithms are most efficient.

Kaan Çağlan, Abdullah Eryüzlü
Developing a Vulnerability Scanner
ZTP(Vulnerability Scanning Program) is a program which scans the devices in the network. It checks the installed programs/packages and report the vulnerable ones, does brute-forcing and report weak passwords, checks for protocol weaknesses (like anonymous login is enabled on the FTP server) and checks Windows registry to report any vulnerabilities.

Mehmet Arda Aksoydan
Development of an Agile Software Product Line for Board Games using Behavior Driven Development Methodology
A product line approach to software production facilitates tailoring products to individual customers. In this project, a software product line is developed using feature oriented approach. The feature-oriented approach makes features explicit in requirements, design, code, testing, and so forth across the entire life cycle of software.

Ali Görkem Yalçın, Ben-Malik Tchamalam, İrem Kaya
Urla Events Mobile Application
The Evde’iz mobile application enables users to get informed about and have access to all kinds of activities from entertainment to sports, from concerts to personal development in a short time. Additionally, it provides access to all information related to the epidemic during the epidemic process we are in. Website available at https://evdeiz.com

Ahmet Şemsettin Özdemirden, Elif Duran, Mustafa Berkay Özkan
MeepleEngine: A Software Product Line For Online Board Games With Unity3D and Spring Boot 2
The “Meeple Engine” is a software-development product designed for users to build a board game, define the rules and design the game logic with respect to core capability, from scratch rapidly. It has three types of users; designers, players and developers. The designers must be aware of the engine document and write a game description file. Players play developed games. Developers add functionalities and maintains core.

Yağız Yiğit Ölmez
Designing and implementing a web based order management software for restaurants

Mustafa Savran, Kaan Uyar, Sulhattin Taşlıtepe
Developing reinforcement learning model for game
In this project , our aim is to build a reinforcement learning models that try to solve environment. The game is target finding game that the agent try to find the shortest way to reach the target and it is written in C++ by using OpenGL library. Model is implemented in python and used pytorch library for training. To make training fast , the model is trained on Google Colab (Tesla K80).

Utku Savaş
Prediction For Pedestrian Crosswalk Behavior
In this project, I tried to classify pedestrians whether he/she going to cross or not. To achieve this, first I detect pedestrians using a detection model and then save their behaviors in a video format. I passed this video as a parameter to another deep learning model. The second deep learning model takes this video and extract several frames from it. Then according to these frames, it classifies the behavior of pedestrians crossing or not.

Nuri Urtekin, Ayşenur Erdem, Berk Bildirici
Tracking Application for Alzheimer Patients
In alzheimer diseases, social skills, behaviors, and logical thinking are adversely affected over time. In many cases, patients leave their premises unnoticed and cannot find the way back. Based on these problems, we decided to produce a system through a mobile application. The application will record the transmitter’s routine (Safe Zone) and contact with the related person if an abnormal condition occurs.

Niyazi Berber, Sevim Tuğba Değirmencioğlu,
Technical Publishing Application
Our software solution is a Techical Publishing Application. The users who interested with technical topic like as software development, technology etc. can read writings about this topic and users can publish content on mobile phone. This software solution makes to create content efficient and easy. We used JavaScript technologies that is responsive in order to use cross-platform mobile development concept, and on the backend side we use express.js, node.js technologies.

Betül Aksoy, Özge Aydın
Quick Recipe Application
The aim of this project is to contribute to the solution of a big and destructive problem such as food waste, and also to assist those who do not know how to cook or learn to cook. Recipes on the internet mostly include the ingredients that are not at home and sometimes very difficult to find. With this application, cooking will become easier and the users of the application will not need to think much about what to cook.

Özge Akat, Işıl Pınarbaşı
Developing a Mobile Application for Stray Animal Emergency
Petmergency is a stray animal emergency application in which you can find first aid information about common emergencies like choking, bleeding, poisoning, etc. or find and contact a veterinary when you encounter a stray animal that needs help. There are two sign-in options in the application: Animal lover and veterinary. Animal lovers can create an emergency appeal and veterinaries can accept the animal if he/she is available.

İrem Erdil, Egehan Karaköse
Developing a Mobile Application for Plant Classification
Mobile application, that use plant’s leaves for classification. The application is able to classify the plant using the picture, taken by the camera of the mobile device or from the device’s gallery. The application that you can use on both iOS and Android platforms, a total of 110 plant leaves are classified. After learning the photo you took from your gallery or the type of plant photo you took instantly, you can access the information of this plant.

Furkan Emre Yılmaz
Lane Assistance and Driving Insight for Drivers / Application
The project aims to design and built a simple lane assistance system. This system will be able to detect lane lines for the driven lane and also the surrounding cars and their respective positions around it. The information which is extracted by the lane assistance system will be displayed in a simple driving computer. The driving computer will be capable of displaying lane lines, vehicles, and the speed of the car.

Yiğit Can Türk, Mehmet Ali Ergül, Seymur Babaşlı
Real-Time Big Data Analytics
By using incremental learning you can create and train models that use streaming big data. But, there are also a lot of successful projects that is created with classical approach. In this study, our main goal is to compare classical and incremental approach in machine learning, determining the advantages and disadvantages of these two different approaches and finally how the number of data affects the success of the model.

Gözde Durmaz
Enhancing a Cybersecurity for IoT Communication
In order to solve IoT security issues, blockchain technology and Json Web Token (JWT) combined in my project. Since blockchain technology and JWT are based on cryptography, their integration into IoT devices provides security. In addition, blockchain technology provides consensus algorithm when IoT devices make a critical decision. The aim of this academic project is to create safer environment for IoT devices by combining these two technologies.

Nur Khaled Aziz, Can Berk Kandemir
Musicity: A Mobile Application for Booking Music Studios and Practice Rooms
We aim to create a mobile application that allows the musicians to book a music studio or practice room nearby or in any city that the user wanted. The user can book a music studio that is available at a certain date and which musical instruments that is available to play. By having this app, the user can also see which music studio/room has a good user rating and review.

Yiğit Can Özkan
Development of a Mobile Phone Controlled Computer Game
Karty is a Co-op 2 to 4 people multiplayer kart racing game for computers. The game is made using Unity Engine. Using local area wireless networks, game is played with android phones, where the phones work as controllers with the help of accelerator sensors built into most smart phones. With more cosmetics, racing tracks, environment styles, boosts, special power-ups, online rankings and online match-makings game has much planned features.

Furkan Şahin, Hüseyin Berk Gök, Berkay Karakoç
Company Insight based on Social Media
Company Insighter aims to build a tool that narrows down the huge search field over the internet for a specific company, therefore, making it easy to analyze their digital footprint over the social media. The main feature of this project is to create a dashboard containing the analysis of tweets on Twitter about a target company. We are aiming to do the analysis by fetching the raw data and make sense out of it with NLP tools like sentiment analysis.

Ogün Çetiner
Stock Market and Currency Exchange Rates Web Application
Nowadays, the stock market and currency exchange information are very easy to access. But it is labor-intensive to invest by constantly analyzing them. For this reason, this project aims to show stock market and foreign exchange rates information, which is normally shown separately, in a single roof, and to save users time. The user can add and remove the desired tables and follow the data instantly.

Murat Karyağdı
Developing a General Real Time Game Server

Şeref Berk Atik, Merve Bozoğlu
Street Basketball
We are designing a mobile application of an Atari game. This game would involve pixelated old style graphic with fundamentals of the street basketball. It will be played by teams of 2 on each side, one player and three computer bots. Characters are able to shoot through hoops, with score of 2, and they will compete to collect most points. The game will be playable on both Android and IOS devices.

Emre Kılavuz, Emre Süslü
Expense Tracker
People want to know their money where they go and as well as from where money comes in.This is exactly the time when our expense tracker called Tracky comes into play.Tracky can be used to create budget where you can always see the how much assets in the system.Then,all assets that is created can track easily and smoothly.When the user enters some expenses or incomes, related graphs are generated automatically which is really useful because rates of the operations can be tracked easily.