Tübitak is one of the largest technology companies in Turkey.
The project on which I have worked that was supported by Tübitak is called Developing and Evaluating a Web-Based and Augmented Reality-Supported Active Learning System to Develop Science Process Skills of Secondary School 6th Grade Students.
It is simply about developing augmented reality and simulations about the units provided according to Ministry of Education curriculum in Science class and controlling the students' progress by using the web control system.
Some examples of these units are systems of the human body, force and motion, etc.
I was on the development team. I developed augmented reality and simulation scenes and worked on the integration part of the web site of this project.
The project is called Developing the Psychomotor Skills of Children With Autism.
The project is co-funded by the Erasmus+ Programme of the European Union.
It is simply about developing software that helps children with autism develop their psychomotor skills by giving instructions. From the control panel, these children's teachers can watch the progress.
I worked in developing and integrating processes (such as audio integration) for this project.
It is a complete web-based farm management system created for farms to observe animals' situations. It is being used by 500+ farms in Turkey. Farm owners also use the developed tracker, which is used to read RFID earings of animals, and it works integrated with the web system with a Bluetooth connection.
It is also co-funded by the Ministry of Agriculture and Forestry.
I worked in the development team to develop the web pages with our clients' feedback to get fast interactions.
The final target is to create a house plant incubator - a cube that would detect which plant was put in, what diseases it has, and introduce a protocol for its treatment. It will have multiple sensors (temperature, humidity, light, etc.) as well as a way of influencing those factors.
Regarding the scope of our university "Project: Deep Learning" it would be creating software for this tool.
Github, Flask, Azure, Kedro, Torch, Weights and BiasesAVESA
This project is being made for my bachelor thesis to be graduated.
AVESA (Audio-Visual Event Sentiment Analysis) is a project to determine the topic of the video by the most action-packed scenes. A GUI has been created for users to be tried from the link below.
It has also been published as a scientific work that has been prepared in the English language.
Details of the project and to try the model: Avesa
If you are interested in full text (it is also specified in the publications section): Full Text
The application of artificial intelligence in some fields comes with significant constraints (including legal ones). In the case of important financial or medical recommendations, transparency in the operation of the systems is essential. The ability to provide an interpetation or explanation of the model’s operation is also an important factor in building trust in artificial intelligence and user satisfaction.
For this project, we have upgraded FashionGAN project as applying xAI methods to explain the model's decisions.
Generator part has been re-trained with the help of LIME and Saliency Map methods as gradient during the training process has been updated by these methods.
Discriminator part has been applied LIME and Saliency Map methods to explain the model's decisions.
Details of the project will be shared later (Github Repo)
Used Technologies
Lime, Saliency Map, TorchTemporal Data Mining
The aim of the project is to develop/use/extend machine learning algorithms related to processing temporal data (including time series, models with hidden states, etc.) and use them to construct a prototype system for a selected problem/data set.
Details of the project will be shared later (Github Repo)
Used Technologies
TGN, TCGN, Diffusion-based, Pathformer Techniques and moreE-mail Spam Classifier
This is a final project for our Machine Learning class.
This is a simple machine learning project that includes data cleaning, EDA, text preprocessing, model building, evaluation, and improvement stages.
This project is being made at my Erasmus+ mobility.
It is about classifying 5 different species (daisy, tulip, sunflower, dandelion, rose) using deep learning.
Multiple different models have been built to find which deep learning model has performed better (VGG16, MobileNetV2, ResNet50) with preventing overfitting problems.
This project is also being made at my Erasmus+ mobility.
It is about classifying 2 different animals (cat, dog) using deep learning.
Multiple different models have been built to find which optimizer function has been performed better (ADAM, RMSProp, SGD) with preventing overfitting problems.
This project automates the process of scraping, analyzing, and classifying news articles using Flask, Streamlit, PostgreSQL, and Machine Learning models inside Docker containers.
Fetches news from an external API (NewsAPI), saves data into PostgreSQL, cleans & preprocesses the text, classifies news sources using ML models, finds the best-performing model dynamically, saves the trained model for real-time predictions in Streamlit, runs inside Docker for full automation.
Stock Market Sentiment Analysis & Prediction is a machine learning project that predicts stock price movement using Twitter sentiment analysis. The project analyzes tweets, determines sentiment, and predicts if a stock will increase or decrease.
Fetches stock-related tweets, performs sentiment analysis, fetches historical stock prices, applies ML models to predict stock movement, streamlit web app for real-time predictions.
Machine Learning libraries, Yfinance, Tweepy, Streamlit, Google Cloud StorageFace Recognition
It is a project to determine faces by using the Face_Recognition library in Python.
The model finds whether the image contains a face or not. Then, it tries to find the best distance between the images that were saved before and the found face.
It is a project to process the Wine Quality project to get the best accuracy among different machine learning methods with preventing overfitting problems.
Data analysis and visualization methods have been used to prepare the data, and three different machine learning methods have been applied to this data.
You can also reach two assignments that were given before the final project. The final project is Wine Quality Analysis.
Details of the project and the codes: Wine Quality
Used Technologies
Sklearn, Pandas, Seaborn, Numpy, Matplotlib, ScipyRegression Models Using Math
It is a project to use regression models without using any library but its math.
Simple and multilinear regression and polynomial regression models have been built.
Results have been compared with the models that Sklearn libraries have, and the results are so close to each other.
The dataset contains different data of different patients.
Here, measnumber can be seen by each patient, and by clicking the patient ID, depending on ovulation hours, plots can be created and saved as a png file
Mean values of different features can be written to the textboxes.
