A characterization of Sturmian sequences by indistinguishable asymptotic pairs
Authors
Barbieri, S.; Labbé, S.; Starosta, Š.
Year
2021
Published
European Journal of Combinatorics. 2021, 95 ISSN 0195-6698.
Type
Article
Annotation
We give a new characterization of biinfinite Sturmian sequences in terms of indistinguishable asymptotic pairs. Two asymptotic sequences on a full -shift are indistinguishable if the sets of occurrences of every pattern in each sequence coincide up to a finitely supported permutation. This characterization can be seen as an extension to biinfinite sequences of Pirillo’s theorem which characterizes Christoffel words. Furthermore, we provide a full characterization of indistinguishable asymptotic pairs on arbitrary alphabets using substitutions and biinfinite characteristic Sturmian sequences. The proof is based on the well-known notion of derived sequences.
Multi-Agent Path Finding with Mutex Propagation
Authors
Zhang, H.; Li, J.; Surynek, P.; Koenig, S.; Kumar, S.
Year
2020
Published
Proceedings of the Thirtieth International Conference on Automated Planning and Scheduling. Menlo Park: AAAI Press, 2020. p. 323-332. ISSN 2334-0835. ISBN 978-1-57735-824-4.
Type
Proceedings paper
Annotation
Mutex propagation is a form of efficient constraint propagation popularly used in AI planning to tightly approximate the reachable states from a given state. We utilize this idea in the context of Multi-Agent Path Finding (MAPF). When adapted to MAPF, mutex propagation provides stronger constraints for conflict resolution in Conflict-Based Search (CBS), a popular optimal MAPF algorithm, and provides it with the ability to identify and reason with symmetries in MAPF.
New family of symmetric orthogonal polynomials and a solvable model of akinetic spin chain
Authors
Kalvoda, T.; Štampach, F.
Year
2020
Published
Journal of Mathematical Physics. 2020, 61(10), 1-21. ISSN 0022-2488.
Type
Article
Annotation
We study an infinite one-dimensional Ising spin chain where each particle interacts only with its nearest neighbours and is in contact with a heat bath with temperature decaying hyperbolically along the chain. The time evolution of the magnetization (spin expectation value) is governed by a semi-infinite Jacobi matrix. The matrix belongs to a three-parameter family of Jacobi matrices whose spectral problem turns out to be solvable in terms of the basic hypergeometric series. As a consequence, we deduce the essential properties of the corresponding orthogonal polynomials, which seem to be new. Finally, we return to the Ising model and study the time evolution of magnetization and two-spin correlations.
Bounds on the period of the continued fraction after a Möbius transformation
Authors
Year
2020
Published
Journal of Number Theory. 2020, 212 122-172. ISSN 0022-314X.
Type
Article
Annotation
We study Möbius transformations (also known as linear fractional transformations) of quadratic numbers. We construct explicit upper and lower bounds on the period of the continued fraction expansion of a transformed number as a function of the period of the continued fraction expansion of the original number. We provide examples that show that the bound is sharp.
Bifurcations and monodromy of the axially symmetric 1:1:−2 resonance
Authors
Efstathiou, K.; Hanßmann, H.; Marchesiello, A.
Year
2019
Published
Journal of Geometry and Physics. 2019, 146 ISSN 0393-0440.
Type
Article
Annotation
We consider integrable Hamiltonian systems in three degrees of freedom near an elliptic equilibrium in 1:1:−2 resonance. The integrability originates from averaging along the periodic motion of the quadratic part and an imposed rotational symmetry about the vertical axis. Introducing a detuning parameter we find a rich bifurcation diagram, containing three parabolas of Hamiltonian Hopf bifurcations that join at the origin. We describe the monodromy of the resulting ramified 3-torus bundle as variation of the detuning parameter lets the system pass through 1:1:−2 resonance
Discovering predictive ensembles for transfer learning and meta-learning
Authors
Kordík, P.; Frýda, T.; Černý, J.
Year
2018
Published
Machine Learning. 2018, 107(1), 177-207. ISSN 0885-6125.
Type
Article
Annotation
Recent meta-learning approaches are oriented towards algorithm selection, optimization or recommendation of existing algorithms. In this article we show how data-tailored algorithms can be constructed from building blocks on small data sub-samples. Building blocks, typically weak learners, are optimized and evolved into data-tailored hierarchical ensembles. Good-performing algorithms discovered by evolutionary algorithm can be reused on data sets of comparable complexity. Furthermore, these algorithms can be scaled up to model large data sets. We demonstrate how one particular template (simple ensemble of fast sigmoidal regression models) outperforms state-of-the-art approaches on the Airline data set. Evolved hierarchical ensembles can therefore be beneficial as algorithmic building blocks in meta-learning, including meta-learning at scale.
Multitask Learning for Cognitive Sciences Triplet Analysis
Author
Tsimafei Stambrouski
Year
2024
Type
Bachelor thesis
Supervisor
Rodrigo Augusto da Silva Alves, Ph.D.
Reviewers
Mgr. Alexander Kovalenko, Ph.D.
Department
Summary
This bachelor thesis focuses on the analysis of the triplet problem, whose task is to identify the odd object out of three. There are different methodological approaches to solving this problem, which differ in their basic focus - some methods evaluate which two objects are most similar, while others identify the object that is most dissimilar. The aim of this thesis is to combine these two perspectives and analyze the results. To solve the problem, a neural network has been developed using the TensorFlow library in the Python programming language. The actually research showed that the combination of both approaches did not produce better results than the individual methods alone. The main output of the work is the explanation of how the combination of opposing views affects the final choice of the odd object. For instance, in the set of (Car, Dog and House) one item should be selected as the odd out. House and Car are the most similar or Car is the odd-one-out?
Robojudge - analysis of court judgments
Author
Miroslav Řehounek
Year
2024
Type
Bachelor thesis
Supervisor
Mgr. Adam Szabó
Reviewers
doc. Ing. Ivan Šimeček, Ph.D.
Department
Summary
The thesis explores how large language models (LLMs) can be used to analyze publically available court rulings of Czech civil courts. The main goal of the thesis was to make these rulings more accessible through the Robojudge web application which allows semantic search of the most similar rulings and presents them in an easy-to-understand way. To select the most appropriate LLM, a custom methodology was devised for comparing such models. This methodology involves evaluating responses from LLMs to legal questions by both human assessors and automatically through another model. Based on this methodology, the most suitable model was then selected. The outcome of the thesis is the aforementioned Robojudge web application that automatically retrieves publicly disclosed court rulings and stores them in a vector database, thus enabling the search of these rulings through natural language. The rulings are retrieved by a scraper that periodically runs itself to maintain the vector database up-to-date. Search results are then presented as concise summaries generated by an LLM. Additionally, users have the option to easily query the content of a specific court ruling, once again through natural language.
System for optimal monitoring of live streams of elephants enclosure
Author
Huu Quy Nguyen
Year
2024
Type
Bachelor thesis
Supervisor
Ing. Jakub Novák
Reviewers
Ing. Magda Friedjungová, Ph.D.
Department
Summary
The thesis focuses on creating a recommender system for optimizing the monitoring of live elephant streams in real time. The work includes a survey of related works focusing on object detection, especially animal detection. The first part of the practical part of the thesis focuses on the collection of data from 8 cameras located in different parts of the elephant complex Elephant valley in Prague Zoo. A dataset of 10205 images with 13034 annotated elephant objects was created. Subsequently, a model based on the YOLOv8 architecture was trained to achieve a mAP of 98 % at an IoU threshold of 0,5 on the test set.
Based on the annotated images, a heatmap was generated using a perspective transformation to show the areas with the highest elephant frequency.
The second part of the practical part focuses on building a decision tree model using the obtained annotated data, which recommends the optimal camera for tracking to the user based on time and day of the week. The resulting model recommends the optimal camera (showing the highest number of elephants) in 62 % of the cases and a camera showing at~least one elephant in 93 % of the cases.
Exploiting Logic Synthesis in SAT-solving
Author
Jan Kimr
Year
2024
Type
Bachelor thesis
Supervisor
doc. Ing. Petr Fišer, Ph.D.
Reviewers
Ing. Robert Hülle, Ph.D.
Department
Summary
This thesis analyzes the possibility of using logic synthesis and optimization to speed up solving satisfiability problem (SAT) instances coming from both the standard benchmarks and different practical applications.
Logic optimization, in principle, influences an instance in two main ways: (1) it reduces its size, and (2) structures making the instance difficult to solve could be dissolved. However, it is necessary to consider both times of synthesis and solving, as speed up in solving achieved by "stronger" synthesis could be outweighed by synthesis time, leading to an overall time increase.
The efficiency of logic synthesis and optimization was evaluated using different instances and syntheses. It can both positively and negatively influence the overall time of solving, which mostly depends on the instance size and used synthesis. Based on the results, recommendations regarding the practical use of synthesis are made.
Artificial Intelligence based Detection and Tracking of Sperm Cells
Author
Jakub Hořenín
Year
2024
Type
Bachelor thesis
Supervisor
Ing. Miroslav Čepek, Ph.D.
Reviewers
Mgr. Alexander Kovalenko, Ph.D.
Department
Summary
This thesis introduces a novel machine learning-based methodology for automated detection and tracking of sperm cells within microscopic video recordings, aiming to elucidate the dynamics and motion patterns of both individual sperm cells and sperm cell bundles. At first, the method identifies sperm cells across successive frames within a video sequence, facilitating the reconstruction of each cell's trajectory over time. Subsequently, I introduce a classification algorithm that distinguishes between single sperm cells, clusters of adjacent cells, sperm cell bundles, and clutter, addressing a gap in existing methodologies. Finally, I employ three conventional metrics for velocity assessment: Straight Line Velocity (VSL), Average Path Velocity (VAP), and Curvilinear velocity (VCL), to quantify the movement speed of both individual sperm cells and bundles. The approach represents a significant advancement in the automated analysis of sperm motility and aggregation phenomena, providing a robust tool for researchers to study sperm behavior with enhanced accuracy and efficiency. A web-based user interface has been created, and the latest version of the program utilizing this methodology is publicly available at https://apps.datalab.fit.cvut.cz/sperm_tracking/ with source code publicly available at gitlab: https://gitlab.fit.cvut.cz/horenjak/sperm_cell_tracking_app/.
Kubaturní Kalmanův filtr v inženýrské praxi
Author
Petr Fiedler
Year
2024
Type
Bachelor thesis
Supervisor
doc. Ing. Kamil Dedecius, Ph.D.
Reviewers
Ing. Ondřej Tichý, Ph.D.
Department
Summary
This thesis explores the cubature Kalman filter, an algorithm designed for estimating system states in real-time within nonlinear models. While the filter possesses solid theoretical foundations, the original exposition of this algorithm may present accessibility challenges due to its reliance on advanced mathematical concepts. This thesis aims to present its derivation comprehensively, making it more accessible while explaining its core mathematical principles. Additionally, the cubature Kalman filter is compared with other algorithms for nonlinear filtering to highlight its unique features. Through practical examples, this thesis demonstrates the filter's effectiveness and operation in various scenarios.
Estimation of detection probability in multitarget filters using object advanced image processing techniques
Author
Michal Seibert
Year
2024
Type
Master thesis
Supervisor
doc. Ing. Kamil Dedecius, Ph.D.
Reviewers
Ing. Ondřej Tichý, Ph.D.
Department
Summary
This thesis focuses on the problem of multi-target tracking with noisy measurements in a cluttered environment. Algorithms for tracking multiple targets with the uncertainty in the number of targets and their survival are often sensitive to the correct setting of parameters relative to the environment. One of the key parameters is the detection probability, which is often constant. In this work, a method to estimate this probability for each target at any time step is demonstrated through tracking objects in video footages. Advanced deep-learning algorithms for image processing are utilized to provide targets' measurements. Outputs of these methods are then exploited to calculate the detection probability. To analyse the capability of the proposed method, the Gaussian mixture probability hypothesis density (GM-PHD) filter is used. This filter falls among fundamental random finite sets statistics-based ultitarget algorithms. The appropriateness of the proposed method is then discussed with regards to its limitations and possible future improvements.
Graph-Based Fraud Detection in Recommender Systems
Author
Daniel Bohuněk
Year
2024
Type
Master thesis
Supervisor
Rodrigo Augusto da Silva Alves, Ph.D.
Reviewers
doc. Ing. Pavel Kordík, Ph.D.
Department
Summary
Fraudsters attempt to camouflage their behavior to remain undetected. This can make it challenging to design models capable of reliably discovering them, as negative samples may be contaminated with hidden positives. Existing research has shown that taking advantage of relationships between instances using graph convolution improves the detection ability. This work proposes a siamese graph neural network that can be trained in a semi-supervised fashion using a small set of known fraudsters. It shows improved performance over existing methods and increased resilience against camouflaged fraudsters.
Machine Learning Techniques for Laser-Plasma Acceleration Optimization
Author
Matěj Jech
Year
2024
Type
Master thesis
Supervisor
Mgr. Alexander Kovalenko, Ph.D.
Reviewers
doc. Ing. Ivan Šimeček, Ph.D.
Department
Summary
The thesis deals with the analysis of data from the laser-plasma particle accelerator in collaboration with the scientific institution ELI Beamlines. In the scope of the work, a data pre-processing process was designed and a generative model simulating the course of physics experiments was developed. The model is conditioned on a vector of experimental parameters and generates image data showing the energy spectrum of the accelerated electron beam. The developed model can be used as a partial substitute for real experiments, which are costly in terms of time and finances. It can also be used as a simulation of real experiments for various optimization methods. This thesis defines the process of training and testing candidate models with three different architectures and based on four hyperparameters. The resulting model can generate data at a rate of 1.8 images per second and has been evaluated based on a number of metrics, including the expert opinion of scientists, as a trustworthy tool to simulate the electron acceleration process.
Explainability in deep learning-based medical image analysis
Author
Martin Lank
Year
2024
Type
Master thesis
Supervisor
Ing. Magda Friedjungová, Ph.D.
Reviewers
Ing. Daniel Vašata, Ph.D.
Department
Summary
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.
Maps of elections
Author
Jitka Mertlová
Year
2024
Type
Bachelor thesis
Supervisor
doc. RNDr. Dušan Knop, Ph.D.
Reviewers
Ing. Daniel Vašata, Ph.D.
Department
Summary
The maps of elections framework is a methodology for visualizing and analyzing election datasets. So far, this framework was restricted to collections of election datasets that have equal numbers of candidates. In this thesis, we extend this framework to be able to deal with datasets containing different numbers of candidates. We do this in two ways: we extend an already existing positionwise distance, and we also develop the idea and implementation of a feature distance. We test the new distances on synthetic data generated using the Mapel package.
Ontologies in Recommender Systems
Author
Ing. Stanislav Kuznetsov
Year
2023
Type
Dissertation thesis
Supervisor
doc. Ing. Pavel Kordík, Ph.D.
Reviewers
prof. dr hab. inz. Krzysztof Goczyla
Assoc. Prof. Aleksandra Klašnja Milicevic, Ph.D.
Mgr. Ladislav Peška, Ph.D.
Assoc. Prof. Aleksandra Klašnja Milicevic, Ph.D.
Mgr. Ladislav Peška, Ph.D.
Detection and removal of watermarks from image data
Author
Tomáš Halama
Year
2023
Type
Master thesis
Supervisor
Ing. Miroslav Čepek, Ph.D.
Reviewers
Ing. Magda Friedjungová, Ph.D.
Department
Summary
Digital image watermarking is a widely used technique for protecting intellectual property or authenticating digital media, but it can negatively impact image quality and usability. This motivates the need for removing watermarks from images, and deep learning presents a potential solution. This thesis develops a deep learning method for watermark removal, including a survey of existing techniques and the proposal of a novel architecture. The method's performance is evaluated in terms of watermark detection accuracy and image reconstruction quality.
Short-Term Precipitation Forecasting from Satellite Data Using Machine Learning
Author
Jiří Pihrt
Year
2023
Type
Master thesis
Supervisor
Mgr. Petr Šimánek
Reviewers
doc. Ing. Kamil Dedecius, Ph.D.
Department
Summary
Geostationary meteorological satellites are a source of global and frequent weather observations, but they do not directly observe precipitation. We research existing methods for inferring and forecasting rainfall from satellite data. The aim of this thesis is to predict high resolution precipitation radar observations up to 8 hours ahead from larger context but lower resolution multi-spectral geostationary satellite images. We develop a novel deep learning model for this task, utilizing the U-Net and PhyDNet neural networks. We name it WeatherFusionNet, as it fuses three different ways to process the satellite data; predicting future satellite images, estimating precipitation in the input sequence, and using the input sequence directly. To train and test it on real data, we participate in the NeurIPS Weather4cast 2022 competition, which provides spatially and temporally aligned satellite imagery and target precipitation radar data. WeatherFusionNet achieved first place in the Core challenge of the competition. We further experiment with several different models, try including static data in the input, and compare our model with a direct radar-to-radar model.
Constraint Programming in Scheduling for Garage
Author
Petr Švec
Year
2023
Type
Master thesis
Supervisor
prof. RNDr. Pavel Surynek, Ph.D.
Reviewers
Ing. Daniel Vašata, Ph.D.
Department
Summary
This thesis deals with the implementation of a system for scheduling workshop work in a car repair shop. The thesis analyses the customer requirements, based on which it defines a model. Model is using constraint programming. Based on the proposed model, we implemented the solver using the choco library. We validated the solver on synthetic and practice motivated data. Various properties of the generated solution for the test instances were measured.
The focus was on universal use with maximum parameterization for the needs of individual clients and integrations.
Scalable Gaussian processes for surrogate modelling in Bayesian optimization
Author
Iveta Šárfyová
Year
2023
Type
Master thesis
Supervisor
Ing. Jiří Vošmik
Reviewers
Ing. Daniel Vašata, Ph.D.
Department
Summary
Bayesian optimisation is a global optimisation method suitable for finding extrema of expensive-to-evaluate black-box objective functions. Gaussian Processes are frequently used as models for approximating such functions. However, their cubic time complexity limits their deployment to applications in small-data regimes. This thesis provides an overview of state-of-the-art scalable Gaussian Processes for regression. The experiments performed within this work deal with tasks of regression and Bayesian optimisation, both utilising several selected Gaussian Process models. Evaluation is done using multiple metrics, some of which are particularly appropriate for probabilistic models. Our results suggest that the same few models consistently outperform the others in both tasks.
Pose and Expression transfer between face images
Author
Petr Jahoda
Year
2023
Type
Bachelor thesis
Supervisor
Ing. Jan Čech, Ph.D.
Reviewers
Ing. Magda Friedjungová, Ph.D.
Department
Summary
This thesis proposes a method for pose and expression transfer between face
images. Given a source and target face portrait, the designed network pro-
duces an output image where the pose and expression from the source face
image are transferred onto the target identity. The architecture consists of
two encoders and a mapping network that maps the two inputs into the latent
space of StyleGAN2, which generates a high-quality image. The training is
self-supervised without the need for labeled data. Our method achieves close
to real-time image generation while also enabling the synthesis of random
identities with independently controllable pose and expression.
Automatic poetic metre detection
Author
Kristýna Klesnilová
Year
2022
Type
Bachelor thesis
Supervisor
Ing. Karel Klouda, Ph.D.
Reviewers
Ing. Magda Friedjungová, Ph.D.
Department
Summary
This work is devoted to automatic metrical analysis of Czech syllabotonic verse metrically tagged inside a large poetic corpus - the Corpus of Czech Verse. First, it reimplements the existing data-driven approach used by a program called KVĚTA. Later, it models the problem as a sequence tagging task and solves it using machine learning. The BiLSTM-CRF model is used, representing the current state of the art for many sequence tagging tasks. Many different input configurations are tested. In all experiments, the inputted syllables or word tokens are represented by Word2Vec word embeddings trained on training data. The results are evaluated by computing three different accuracies of the predictions: syllable-level accuracy, line-level accuracy, and poem-level accuracy. It is shown that using BiLSTM-CRF represents a great success. With the best input configurations, it produces better results than the KVĚTA reimplementation, with predictions achieving 99.61% syllable accuracy, 98.86% line accuracy, and 90.40% poem accuracy. The most interesting finding is that the best results are obtained by inputting sequences representing whole poems instead of individual poem lines.
Bilingual search in documents
Author
Lukáš Rynt
Year
2022
Type
Bachelor thesis
Supervisor
Ing. David Bernhauer, Ph.D.
Reviewers
prof. Dr. Ing. Petr Kroha, CSc.
Department
Summary
This thesis is concerned with the research of information retrieval models, state-of-the-art word embedding techniques and their possible use for multilingual retrieval. Modern approaches to multilingual retrieval that build on word embedding techniques usually work with some transformation that converts word representations from one language to another. The aim of this thesis is to investigate a model that does not work with this transformation and instead directly extracts dependencies between translations. A prototype web search engine should then be built on top of this model.
The work has met all expectations and the resulting model was able to represent bilingual translations directly without using any transformation. This was achieved using parallel translated European Union documents, which were linked at paragraph level for both languages. The prototype search engine then operated based on the learned word representations derived from this model.
Hierarchical Control of Swarms during Evacuation
Author
Kristýna Janovská
Year
2022
Type
Bachelor thesis
Supervisor
prof. RNDr. Pavel Surynek, Ph.D.
Reviewers
doc. Ing. Pavel Hrabák, Ph.D.
Department
Summary
In this work I propose a hierarchical system of multi-agent coordination for simulating evacuations. I distinguish between two types of agents. Leaders communicate between themselves and escort their swarms to a safe location, while followers follow their leader. For this multi-agent path finding algorithms are used.
I propose multiple models varying both in leaders' behavior towards their respective swarms and followers' behavior considering their attempts at individual evacuation.
In the work I perform experiments, whose results will show how parameters of agents' behavior influence the success of an evacuation. Results of these experiments will point out advantages of communication between leaders, problems, that might occur during an evacuation and their connection to agents' inadequate behavior.
Flow modelling around airfoil with graph neural networks
Author
David Horský
Year
2022
Type
Bachelor thesis
Supervisor
Mgr. Vojtěch Rybář
Reviewers
Ing. Daniel Vašata, Ph.D.
Department
Summary
In this thesis we reviewed uses of machine learning in computational fluid
dynamics. We then implemented a state-of-the-art graph neural network to
simulate the flow around an airfoil in 2D. We train the model at lower speeds
and angles of attack and then extrapolate to higher ones. We trained a model
that extrapolates with a small precision error and remains stable on long
rollouts.
Design of a camera system for scanning a tire tread
Author
Daniel Bohuněk
Year
2022
Type
Bachelor thesis
Supervisor
doc. RNDr. Ing. Marcel Jiřina, Ph.D.
Reviewers
Ing. Tomáš Kalvoda, Ph.D.
Department
Summary
The objective of this bachelor's thesis is to design a camera system for scanning tire treads of passing vehicles in a standardized form.
After researching existing solutions for scanning tire treads a new camera system is proposed, which captures and unwraps the tire tread of vehicles stopped at a traffic light. A convolutional neural network is used to find the location of the tire in an image.
The camera system was tested in real traffic at multiple traffic lights. 78 % of vehicle tires are captured correctly. The implemented algorithms correctly process the tire tread 89 % of time.
This new camera system can be used to collect large amount of tire tread samples, which might otherwise be impractical and expensive to obtain. This data is useful for example in forensic analysis, to pair a tire tread with tracks left on the ground.
Optimization of the use of traditional segmentation algorithms for defect detection tasks in industry
Author
Jiří Szkandera
Year
2022
Type
Bachelor thesis
Supervisor
Ing. Jakub Novák
Reviewers
Ing. Miroslav Čepek, Ph.D.
Department
Summary
The paper compares algorithms designed for image segmentation. Three algorithms belonging to the category of superpixels (felzenszwalb, SLIC and quickshift), two representatives of active contour models (snakes and level sets), random walker, region adjacency graphs and Otsu thresholding are compared. For this purpose, defects appearing in the industry are described. A more general categorization splitting defects into groups is made. Subsequently, two defects are selected from each category. A suitable combination of algorithm parameters is found on the first defect. Different preprocessing and color representation of the image are accounted for when the search is carried out. With the found parameters, a segmentation of the second defect is performed. Thus, the generalization capability and the suitability of the algorithm for the defects of the given category are tested. The success of the segmentation is measured by the IOU metric. The more successful algorithms achieved an average IOU measured over all images of a single defect of 90 %. In generalization test an average IOU of 53 % was achieved in some cases.
Anomaly detection on the CERN data centre monitoring data
Author
Antonín Dvořák
Year
2022
Type
Master thesis
Supervisor
Mgr. Alexander Kovalenko, Ph.D.
Reviewers
doc. Ing. Kamil Dedecius, Ph.D.
Department
Summary
One of the many tasks of CERN cloud service operators is to make sure that the desired computational power is delivered to all users of the scientific community. This task is accomplished by carefully setting threshold-based alarming on top of the infrastructure performance time series metrics.
In order to maximize the efficiency of the cloud infrastructure and to reduce the monitoring effort for service operators, we have developed a fully automated Anomaly Detection System that leverages unsupervised machine learning methods for time series metrics. Moreover, adopting ensemble methods, we combine traditional (Isolation forest) and deep learning (Gated recurrent unit/Long short-term memory Autoencoders) approaches.
This work presents a description of the CERN monitoring infrastructure, problem formulation, design of the Anomaly Detection Pipeline, description of used models, creation of the dataset and performance of the implemented models compared to the performance of the Current Alarming System.
The Study of Linear Self-Attention Mechanism in Transformer
Author
Uladzislau Yorsh
Year
2022
Type
Master thesis
Supervisor
Mgr. Alexander Kovalenko, Ph.D.
Reviewers
doc. Ing. Pavel Kordík, Ph.D.
Department
Summary
As the quadratic complexity of an attention mechanism in the Transformer architecture places a high demand on processing long sequences, the goal of this research is to explore possibilities of linear attention in Transformer-like architecture and implement new methods.
Lazy Compilation in Classical Planning
Author
Zuzana Fílová
Year
2022
Type
Master thesis
Supervisor
prof. RNDr. Pavel Surynek, Ph.D.
Reviewers
Ing. Daniel Vašata, Ph.D.
Department
Summary
The subject of this diploma thesis is focused on a lazy compilation in classical planning. The theoretical part summarizes the basics of classical planning. Key concepts of the classical representation of planning problems are defined and basic planning algorithms are presented, in particular, the search in the planning state space and techniques using the planning graph. The compilation of the planning problem into the propositional satisfiability problem (SAT) is discussed at the end of this section.
Based on the obtained knowledge, a new method for lazy compilation of planning problems into SAT has been proposed. Different from the classical compilation, in this method the propositional formula is gradually created and modified. As part of the practical part of the work, a planner was implemented using two compilation variants - the proposed method for lazy compilation and classical compilation.
The planner was tested on planning problems from the International Planning Competition (IPC). The experiments focused on evaluating the success of the planner based on lazy compilation and comparing the results with the planner using the classical compilation method. A total of 79 problems of varying difficulty from four domains were used, of which the lazy planner was able to solve 63 faster than the classical planner. The performed experiments pointed out the advantages and possible disadvantages of lazy compilation. The results of the experiments indicate that the use of lazy compilation has the potential to improve the performance of the planner.
Improving deep learning precipitation nowcasting by using prior knowledge
Author
Matej Choma
Year
2022
Type
Master thesis
Supervisor
Mgr. Petr Šimánek
Reviewers
Mgr. Petr Novák, Ph.D.
Department
Summary
Deep learning methods dominate short-term high-resolution precipitation nowcasting in terms of prediction error. However, their operational usability is limited by difficulties explaining dynamics behind the predictions, which are smoothed out and missing the high-frequency features due to optimizing for mean error loss functions. This thesis summarizes our progress in addressing these issues.
Firstly, we present Intensity Classification Loss to improve the prediction of severe rainfall. The model is trained to predict the probability of precipitation with an intensity over 40 dBZ as a secondary output, which is compared to binary ground truth. Experiments have shown that this approach helps predict severe rainfall but does not predict precipitation with higher intensities than the selected threshold.
Secondly, we experiment with hand-engineering of the advection-diffusion differential equation into a PhyCell to introduce more accurate physical prior to a PhyDNet model that disentangles physical and residual dynamics. Results indicate that while PhyCell can learn the intended dynamics, training of PhyDNet remains driven by loss optimization, resulting in a model with the same prediction capabilities.
Power line vegetation management using UAV images
Author
Radek Ježek
Year
2022
Type
Master thesis
Supervisor
Ing. Lukáš Brchl
Reviewers
doc. Ing. Štěpán Starosta, Ph.D.
Department
Summary
The electric utility companies spend large amounts of money and effort every year to ensure the safe and uninterrupted operation of the electric power infrastructure. The most common source of outages is vegetation damaging power lines, for example, fallen trees. For this reason, companies perform regular inspections and maintenance of power line corridors, especially in forests and densely vegetated areas, creating a high demand for inexpensive and highly automated methods of power line corridor surveys. This work aims to create a robust algorithm for automatic detection of vegetation encroachment in the power line corridor using an Unmanned Aerial Vehicle (UAV), the techniques of photogrammetry, and computer vision. The study will cover the workflow for power line corridor inspection from comprehensive guidelines for data acquisition through power line 3D reconstruction to vegetation encroachment detection and visualization of the results.
Predictor Factory: Learning from Relational Data
Author
Ing. Jan Motl
Year
2022
Type
Dissertation thesis
Supervisor
doc. Ing. Pavel Kordík, Ph.D.
Reviewers
Prof. Abdullah Uz Tansel
doc. Mgr. Martin Nečaský, Ph.D.
doc. Ing. Tomáš Kliegr, Ph.D.
doc. Mgr. Martin Nečaský, Ph.D.
doc. Ing. Tomáš Kliegr, Ph.D.
Knowledge Extraction from Multimedia Content
Author
Ing. Petr Pulc
Year
2022
Type
Dissertation thesis
Supervisor
prof. Ing. RNDr. Martin Holeňa, CSc.
Reviewers
prof. Irina Perfiljeva, CSc., dr. h. c., prof. h. c.
Assoc. prof. Neeta Nain, Ph.D.
doc. Ing. Karel Zimmermann, Ph.D.
Assoc. prof. Neeta Nain, Ph.D.
doc. Ing. Karel Zimmermann, Ph.D.
Road Quality Classification
Author
Martin Lank
Year
2021
Type
Bachelor thesis
Supervisor
Ing. Magda Friedjungová, Ph.D.
Reviewers
Ing. Daniel Vašata, Ph.D.
Department
Summary
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.
Artificial Intelligence Methods for Interior Design and Furnishing
Author
Eliška Svobodová
Year
2021
Type
Bachelor thesis
Supervisor
Ing. Mgr. Ladislava Smítková Janků, Ph.D.
Reviewers
prof. RNDr. Tomáš Skopal, Ph.D.
Department
Summary
This thesis contributes to automated interior design research by designing and implementing a new system. The design process is divided into the planning of the functional zones using simulated annealing and arranging the furniture with a genetic algorithm. The system can fulfill the user's requirements on the room's shape, functions, and used furniture. The experiments show the ability of the system to design an interior of rooms with varying shapes and selection of furniture.
Embedding interactive elements and multimedia into PDF files using TeX
Author
Michal Vlasák
Year
2021
Type
Bachelor thesis
Supervisor
RNDr. Petr Olšák
Department
Summary
This bachelor thesis concerns itself with the area of interactive features and multimedia in PDF files, especially in connection with TeX. Apart from the analysis and discussion of what PDF standard offers, or what is implemented by existing TeX packages, the focus was also practical--testing what really works in PDF viewers available today.
It turns out, that PDF offers many interactive and multimedia possibilities. The support in PDF viewers is however strongly lacking. The exception is the de-facto reference viewer Acrobat and a viewer strongly influenced by it--Foxit. However, there are open-source viewers (Evince, Okular) whose use in some areas may be completely satisfactory.
The gained knowledge was used to create a package for a new TeX format OpTeX. The package implements those interactive features and multimedia capabilities that have sense in the context of TeX and also work in practice. The package is publicly and freely available.
In the area of multimedia, the resulting package offers the possibility to insert audio, video, and 3D art. In the area of interactive features, it for example complexly handles actions, trigger events, or transitions.
Measurement of dimensions and shapes of costume jewelry diamonds
Author
Justýna Frommová
Year
2021
Type
Bachelor thesis
Supervisor
Ing. Jakub Novák
Reviewers
Ing. Mgr. Ladislava Smítková Janků, Ph.D.
Department
Summary
The focus of this thesis is on the size measurement of jewelry stones based on computer vision and image processing. The quality of jewelry stones is inspected by the size of its circumscribed and inscribed circle, the size of the circumscribed circle of the diamond's table, level of polish and center offset. Three acquisition systems and four algorithms are presented. The acquisition system is composed of a monochromatic camera, a telecentric lens, a coaxial light and a specific bar diffuser and ring diffuser. Algorithms are implemented using the Canny-Otsu detector, Otsu thresholding and watershed segmentation. The accuracy of the measurements is 0,0098 mm, size deviation related to the position in the image oscillates within a range of 0,0028 mm and 0,0040 mm.
Compilation of Multi-Agent Collective Construction in the Minecraft Game
Author
Martin Rameš
Year
2021
Type
Bachelor thesis
Supervisor
prof. RNDr. Pavel Surynek, Ph.D.
Reviewers
Dr. techn. Ing. Jan Legerský
Department
Summary
This bachelor thesis studies current exact approaches to multi-agent collective construction problem, with emphasis on three-dimensional structures, built by agents repositioning passive blocks on a grid under the condition of gravity. A generalization of current fastest exact model is proposed, using mixed integer linear programming, to accommodate varying durations of agent actions. An application of the proposed model is used in conjunction with a solver to exactly optimize construction plan of user entered structures. The result is visualized in Minecraft, using program built upon project Malmo API. A series of experiments is performed on several small instances to measure relative construction makespan reduction, relative to the instances with one-timestep actions. Results show significant makespan reduction at action durations used for visualization in Minecraft.
Optical rods defects detection and classification
Author
Matěj Latka
Year
2021
Type
Bachelor thesis
Supervisor
Ing. Jakub Novák
Reviewers
Ing. Magda Friedjungová, Ph.D.
Department
Summary
This thesis focuses on the automated detection and classification of defects on glass rods. Previous solutions of similar problems are analysed and a new approach is proposed. Four camera systems using advanced optical components and lighting are designed. Also, analysed defect detection and classification methods are modified and extended. One of them using a model based on the modern Faster R-CNN architecture is able to detect 83 % of defects and to correctly classify 77 % of them.
Hardware tool for precise targeting of the camera view in the room
Author
Zuzana Jiránková
Year
2021
Type
Master thesis
Supervisor
Ing. Jakub Novák
Reviewers
Dr.-Ing. Martin Novotný
Department
Summary
The aim of the work is to create a hardware system to focus the view of industrial camera in the sense of getting higher resolution in area of interest. A solution to use two cameras was chosen. The first one is an overview camera and the second one is a detail camera, which gaze is aimed by a mirror tilted in two axis. The created solution enables to aim to a wanted object from an overview camera and retrieve its detail in higher resolution by the detail camera. The main result of the work is the created hardware system and its firmware to control tilting of the mirror.
Sample interactive graphic calculation output
Author
Dominika Králiková
Year
2021
Type
Master thesis
Supervisor
doc. Ing. Štěpán Starosta, Ph.D.
Reviewers
doc. Ing. Ivan Šimeček, Ph.D.
Department
Summary
This diploma thesis focuses on the selection of appropriate technologies for the creation of interactive graphical visualizations of calculation output and on implementation of model examples designated to support the teaching of mathematical subjects with the use of proposed solution. First part of the thesis is dedicated to the analysis of available rendering systems. Subsequently, a list of model examples suitable for interactive demonstrations during a lecture is assembled. Afterwards the thesis focuses on the selection of a suitable rendering system, on the design and the implementation of a web application containing individual model examples, on the testing of this application and on its trial deployment. The output of this thesis is a finished web application implemented with the use of the Plotly library, which includes defined model examples such as function approximation with the help of Taylor polynomial or visualization of the method of Lagrange multipliers. The application supports easy creation of new visualizations and their integration with the MARAST teaching support system.
Algorithms for Combinatorics on Words
Author
Martin Rejmon
Year
2021
Type
Master thesis
Supervisor
doc. Ing. Štěpán Starosta, Ph.D.
Reviewers
doc. Ing. Jan Janoušek, Ph.D.
Department
Summary
In this thesis, several algorithms for problems from the mathematical field combinatorics on words are thoroughly explained. The algorithms are also implemented in the free and open-source mathematics software system SageMath with the goal of their eventual integration into the same system. The algorithms include: classifying mortal and bounded letters of a morphism, deciding whether a morphism is injective, finding a simplification of a morphism, finding all infinite repetitions in a D0L system, and finding all factors (up to a certain length) of words of the language of a PD0L system. Furthermore, the problem of deciding whether a morphism of a D0L system is injective on the set of factors of words of the language of the system is discussed. The decidability of this problem is an open question, which is not answered in this thesis, but it is shown why it is nontrivial and where some approaches to solving this problem fail.
Bayesian filtering of state-space models with unknown covariance matrices
Author
Tomáš Vlk
Year
2021
Type
Master thesis
Supervisor
doc. Ing. Kamil Dedecius, Ph.D.
Reviewers
Ing. Ondřej Tichý, Ph.D.
Department
Summary
This thesis explores the problem of distributed Bayesian sequential estimation of unknown state-spacemodels with unknown processes and measurement noise covariance matrices. This is a frequent problem in real-world scenarios, where the information about noise covariance matrices for specific sensors may not be available. The solution proposed in this thesis is built upon the variational Bayesian paradigm, which is used for the estimation of the states, as well as the unknown measurement noise covariance matrix. From performance improvements, the measurements and posterior estimates are shared between the adjacent node in the network. It also shows a way of optimizing the process noise covariance matrix.
Advanced Methods for Asymmetric Heterogeneous Transfer Learning
Author
Ing. Magda Friedjungová
Year
2021
Type
Dissertation thesis
Supervisor
doc. RNDr. Ing. Marcel Jiřina, Ph.D.
Reviewers
Ing. Vojtěch Franc, Ph.D.; prof. Michał Choraś, Ph.D.; dr. inž. Mariusz Topolski