Ing. Magda Friedjungová, Ph.D.

Ing. Magda Friedjungová

Ing. Daniel Vašata, Ph.D.

Department of Applied Mathematics

In this thesis I deal with the creation of balanced groups of students for collaborative learning based on their results in individual lessons. The thesis also consists of an analysis of available data, implementation of several methods and their comparsion. I have chosen four specific methods: K-means, DBSCAN, Hopfield network and a combination of Self-organizing map and agglomerative hierarchical clustering. One of the results of this thesis is a library for the automatic creation of student groups, which can be embedded into e-learning systems that have the required data structure.

Thesis on DSpace

Ing. Magda Friedjungová, Ph.D.

Ing. Mgr. Ladislava Smítková Janků, Ph.D.

Department of Applied Mathematics

The aim of this bachelor thesis is to create a model that predicts the winners of professional MMA matches. The predictions are based on statistics from previous matches. The research carried out examines the available data sources. To the statistics from previous matches are also added predictions by bookmakers in the form of historical odds. A research of already existing works with a similar theme was also carried out. On the basis of the research, several classification models were selected, which were then trained on the preprocessed data and their predictions were compared with each other, including comparison with the predictions of bookmakers. The results of the best models appear to be favourable.

Thesis on DSpace

Ing. Magda Friedjungová, Ph.D.

Ing. Pavel Karol

Department of Software Engineering

This bachelor thesis deals with the design and implementation of user roles with a focus on access to individual data sets based on assigned roles within CTU. The practical part of the work deals with the analysis of the reporting application Reports, informs the reader about the architecture of the CTU data warehouse, which provides data for the application and also informs him about user roles at the university. The practical part, based on the analysis, extends the functionality of the application to work with roles obtained from the Usermap system, when users are automatically assigned to the groups after logging in, depending on their current business roles. Using the groups they have assigned, users can access certain reports and show the charts which are they authorized for. A database with which the application communicates is used to represent groups and roles, and with its help individual roles and groups can be linked so that users with certain roles are automatically added to linked groups. The solution includes the ability to make the report available to specific users, which is manually assigned by the administrator. The benefit of the work is a system of roles enabling the setting of access to individual reports within the application and the extension of their graphical representation.

Thesis on DSpace

Ing. Magda Friedjungová, Ph.D.

Ing. Daniel Vašata, Ph.D.

Department of Applied Mathematics

Automated evaluation of road quality can be helpful to authorities and also road users who seek high-quality roads to maximize their driving pleasure. This thesis proposes a model which classifies road images into five qualitative categories based on overall appearance. We present a new manually annotated dataset, collected from Google Street View. The dataset classes were designed for motorcyclists, but they are also applicable to other road users. We experimented with Convolutions Neural Networks, involving custom architectures and pre-trained networks, such as MobileNet or DenseNet. Also, many experiments with preprocessing methods such as shadow removal or CLAHE. Our proposed classification model uses CLAHE and achieves 71% accuracy on a test set. A visual check showed the model is applicable for its designed purpose despite the modest accuracy since the image data are often controversial and hard to label even for humans.

Thesis on DSpace

Ing. Magda Friedjungová

Ing. Robert Kotlář

Department of Software Engineering

In this bachelor thesis I deal with the issue of control of ETL processes in the data warehouse. The aim of the thesis is to analyze, design and implement a server to control and monitor these processes. It also allows the user to change configuration and view in real time. In addition to starting the loading of the CTU Data Warehouse, the application allows you to run your own recorded ETL. The solution is made as a Python web application with Flask framework.

Thesis on DSpace

Ing. Magda Friedjungová

Ing. Zdeněk Balák

Department of Applied Mathematics

This thesis deals with the design and implementation of a framework for creating knowledge-based diagnostic systems. The framework is designed as a Java project with the JavaFX libraries for the GUI. The solution provides an interface for a knowledge-based system in which the individual modules can be implemented independently on each other. This ensures the proper functioning of the system with a previously unspecified implementation. There are implementations for most of the modules, which may be used in the system or serve as an inspiration. The goal of this thesis is to improve the way students are taught at BI-ZNS lectures at FIT CTU in Prague and to simplify the teaching process for the teachers themselves.

Thesis on DSpace

Ing. Magda Friedjungová, Ph.D.

Ing. Daniel Dombek, Ph.D.

Department of Applied Mathematics

The low study programme completion rate remains a long-term phenomenon among Czech technical universities. Schools therefore seek to provide timely assistance to students who may have difficulties studying. Data analysis and educational data mining may present a way to identify such students. This thesis presents a comprehensive analysis of study records and other associated data of the students of the Faculty of Information Technology, Czech Technical University in Prague. The data warehouse of CTU was used to obtain the data, with the records ranging from the years 2009 to 2021. Machine learning methods were utilised to create the predictive models from the pre-processed data. The created models are tasked with predicting the successful completion of the study programme by individual students and the overall ratio of students successfully completing the current semester. For chosen faculty courses, the students' final grades are also predicted. On top of that, the discussion of the results' applicability in the future and the automation of the prediction process are among the outputs of this thesis.

Thesis on DSpace

Ing. Magda Friedjungová

Ing. Daniel Vašata, Ph.D.

Department of Applied Mathematics

Restoring damaged regions in image data is a relevant and difficult problem, which gets proportionally harder with the severity of the damage. In the last few years we have seen promising progress in tackling this issue using deep learning models. This thesis verifies and compares different approaches to handling the image inpainting problem. Since generative adversarial networks are one of the most inventive and promising architectures, a survey on current methods was performed and two selected methods were reimplemented. Our implementations are compared to other inpainting methods using classification models. The presented results reflect the influence of damage type and damage severity on the ability of each of the considered methods to successfully inpaint a damaged image.

Thesis on DSpace

Ing. Magda Friedjungová, Ph.D.

Ing. Karel Klouda, Ph.D.

Department of Applied Mathematics

This bachelor thesis deals with the visualization process of clustering methods with emphasis on the application in natural language processing, which are used to solve the unsupervised clustering task in order to perform a topic modeling on a corpus of Czech verse. The paper presents both the clustering methods themselves and possible ways of visualizing them. These methods are further applied to embedded corpus of Czech verse and compared with the help of visualizations.

Thesis on DSpace

Ing. Magda Friedjungová, Ph.D.

Ing. Jakub Novák

Department of Applied Mathematics

This thesis deals with the improvement of faculty candy dispenser. Various methods for face and smile detection in video are described and implemented. The researched methods include, for example, convolutional neural networks, the Viola-Jones algorithm or HOG. The methods are compared on publicly available datasets. A neural network from the Mediapipe library was chosen as the most suitable method for face detection, which achieves better accuracy and speed than the method in the original implementation. The HOG method was chosen for smile detection, which outperformed the convolutional neural network of the original implementation. The resulting methods were deployed on the production environment of the candy dispenser and it is possible to use them.

Thesis on DSpace

Ing. Magda Friedjungová

Ing. Tomáš Kalvoda, Ph.D.

Department of Applied Mathematics

Many different methods have been suggested to colorize images yet. In this thesis, I try to implement a fully automatic image colorization using generative adversarial networks - GANs. They have shown promising results in generating various kinds of data, including images. I adopt two different modifications of GANs - DCGAN and CycleGAN. These two methods are compared, and results are evaluated using the most common metrics used for this problem. Example images are provided as well.

Thesis on DSpace

Ing. Magda Friedjungová

Ing. Daniel Vašata, Ph.D.

Department of Applied Mathematics

Most labelled real-world data is not uniformly distributed within classes, which can have a severe impact on the prediction quality of classification models. A general approach is to overcome this issue by modifying the original data to restore the balance of the classes. This thesis deals with balancing image datasets by data augmentation using generative adversarial neural networks. The primary focus is on generating images of underrepresented classes in imbalanced datasets, which is a process known as class balancing. The aim of this thesis is to analyse and compare data augmentation techniques including standard methods, generative adversarial networks and autoencoders. Evaluation is done using classifiers trained on the original, unbalanced and augmented datasets. The results achieved suggest how the performance of the methods proportionately deteriorates with increasing imbalance rate and diversity of datasets.

Thesis on DSpace

Ing. Magda Friedjungová, Ph.D.

Ing. Karel Klouda, Ph.D.

Department of Applied Mathematics

This bachelor thesis surveys state-of-the-art methods of topic modeling with an emphasis on their use in topic modeling of poetry. The methods are applied on the Corpus of Czech Verse. This corpus contains 1,305 collections of poetry from the 19th and the beginning of 20th century which are lemmatized and annotated phonetically, morphologically, metrically, and strophically. The thesis evaluates the results of the methods and compares them with each other.

Thesis on DSpace

Ing. Magda Friedjungová

Ing. Daniel Vašata, Ph.D.

Department of Theoretical Computer Science

The aim of the thesis was to create a recommendation system of optional courses for students of the Faculty of Information Technology of CTU. Several recommendation methods have been tried, specifically memory-based collaborative filtering and collaborative filtering based on forward neural network. The simpler, memory-based method was chosen for user testing, as it was better in predicting the student's future mark, which was considered by the author to be an important aspect to receive trust from students. The resulting recommendation system can be further improved, the work provides a detailed analysis of the topic as well as sufficiently pre-processed input data, and any further work in this area is therefore facilitated.

Thesis on DSpace

Ing. Magda Friedjungová

doc. Ing. Robert Pergl, Ph.D.

Department of Software Engineering

This master's thesis describes the design and implementation of a data layer solution for the CTU data warehouse, which is being implemented in development projects (DÚ n. 20 a n. 46). The thesis describes the theory of business intelligence, data warehouses, data warehouse architectures and database naming conventions. The implementation part consists of the design of a database naming convention for the data warehouse architecture of the CTU data warehouse. It describes the implementation of individual data layers and the design of a central database. This implementation was tested using datamart visualisation in Pentaho BI Server.

Thesis on DSpace

Ing. Magda Friedjungová, Ph.D.

Ing. Daniel Vašata, Ph.D.

Department of Applied Mathematics

Many real-world tabular datasets are imbalanced, which has a negative effect on the quality of the models applied to those data. More ways exist to deal with this problem, and one of them is re-balancing the dataset. This master's thesis presents a review of existing oversampling methods, such as SMOTE, for imbalanced datasets of binary classification problem. Traditional methods, as well as generative-based techniques, are outlined in this work. Moreover, a novel oversampling method called LIT-GAN is presented. LIT-GAN combines an interpolation technique with generative models. Selected ten oversampling methods and the novel method are compared experimentally using classification on frequently used datasets with six different evaluation metrics.

Thesis on DSpace

Ing. Magda Friedjungová, Ph.D.

Ing. Daniel Vašata, Ph.D.

Department of Applied Mathematics

In this work, we apply Grad-CAM++, LayerCAM and SmoothGrad explainability methods to the proposed EfficientNetV2-based convolutional neural network fine-tuned on microscopic histology imaging. The network predicts mean diffusivity (MD) and fractional anisotropy (FA) obtained from diffusion tensor imaging. The aim of the work was to reveal which histology features tend to increase MD and FA. The proposed network achieved more than 98.5% R2 on all train, validation and test sets, surpassing the network proposed in the preceding work by tens of percentage points in R2. Nevertheless, the explainability methods applied to microscopy imaging were less valuable than anticipated. They indicate certain nuclei influence; however, the details about the relationship remain undiscovered.

Thesis on DSpace

Ing. Magda Friedjungová

Ing. Tomáš Kalvoda, Ph.D.

Department of Software Engineering

This work deals with implementation of a component of MARAST educational system in order to analyse data from quizes. The work includes introduction to the domain and to the provided data export that is further subjected to manual analysis.During analysis, methods of mathematical statistics and data mining were used,including methods of unsupervised machine learning like clusternig and anomally detection. Next part of this work contains development of analytical component. Development consists of analysis, design and implementation of the component that was produced as web application. The result of this work is a functional component for data analysis of quizes from MARAST system, component is tested, documented and deployed.

Thesis on DSpace

Ing. Magda Friedjungová, Ph.D.

Ing. Karel Klouda, Ph.D.

Department of Applied Mathematics

A new course called Data visualization will be taught at the Faculty of Information Technologies of the Czech Technical University in Prague. This thesis describes the process of creating study materials of the mentioned course, based on research of similar courses at other universities and analysis of data visualization methods.

Thesis on DSpace