This thesis addresses human pose estimation using neural networks. The first part describes methods based on convolutional neural networks and methods based on transformer architecture, which have been successfully used to solve this problem. In the second part, a model is chosen...

This work deals with the task of action recognition from video recordings. The I3D method was chosen to solve this task. This metod inflates convolutional neural networks so that they can work with spatio-temporal information stored in videos. The ResNet50 model was chosen as ...

The main goal of this diploma thesis is the design and subsequent implementation of manipulators for controlling the autonomous movement of table football players. In the introduction we will focus on the familiarization of mechatronic systems and the connection of the role of syst...

This diploma thesis deals with the problem of optical character recognition (OCR) using neural networks. I am focusing on improving text detection and OCR by fine-tuning an E2E-MLT scene text detector by training it on synthetic data which emulates real data. The model was fin...

The thesis deals with the semi-supervised document classification. The theoretical part covers the basics of machine learning and defines the space for semi-supervised learning methods between supervised and unsupervised learning. Furthermore, sections of semi-supervised learning methods and document class...

This master thesis deals with the use of artificial neural networks for natural language processing. In recent years, models with the Transformer architecture have become widely used in this field of artificial intelligence. It is these models and their structure that are thoroughly...

This thesis deals with design of wheeled inspection robot control system using ROS framework. The task of the robot is to visit autonomously points of interest, which can be defined in the created user interface. The thesis contains a complete description of the used rob...

Spoken dialogue systems offer a possibility of communication between a~person and a computer. The communication can make a life easier not only for handicaped people (with blindess, physical disability) but also for people in their everyday lives, e.g. making a reservation in a res...

Using neural networks with a newly designed structure, the master thesis replaces unreliable expert methods for inserting regulatory sites into promoters. This is the first use of artificial intelligence to modify promoters in this way. Newly structured models of neural networks were cr...

Researchers of the Department of Cybernetics at the University of West Bohemia in Pilsen in cooperation with SpeechTech s.r.o. have developed a system, which automatically subtitles live broadcasts for Czech Television. The aim of this thesis is to extend the system for the "U...

The diploma thesis deals with the monitoring of vibrations of steam turbine blades. Based on the knowledge associated with this issue, software was created for offline analysis of these vibrations. The resulting application was developed with the support of Logic Elements s...

This master thesis was created in order to test algorithms for automatic setting of parameters of industrially used controllers. The control was tested for an algorithm implemented in industrial computers from Siemens and an algorithm developed by REX Controls and implemented in the...

This thesis is creating a Gantry crane Hardware-In-the-Loop simulator by following through the X-In-the-Loop process. It begins with mathematical model derivation and validation and PID controller tuning and input-shaper to cancel oscillations, is tested in Simulink in Model-In-the-Loop configuration, ...

The diploma's thesis focuses on the optimization of the power transfer capacity. The theoretical part of this diploma's thesis describes Market coupling in Europe, ATC and Flow-based approach, Topology and Power Shifting Transformers remedial actions, Optimal Power Flow methods, Evolutionary&#...

The topic of this diploma thesis is the detection of glottal closure instants from the speech signal using machine learning methods. It aims to improve the success of the classification of the initial algorithm, especially by adding new features and finding other suitable methods&#...

The diploma thesis deals with the problem of semantic segmentation of automotive images using a deep neural networks. DeepLabV3+, the state-of-the-art model on Cityscapes dataset, was re-implemented using Keras and TensorFlow frameworks. The model was pretrained on the ImageNet dataset and&...

This diploma thesis deals with state estimation using Point-Mass Filter. Mainly two topics are discussed; first, a two novel designs of unequally located grid points resulting in improved estimation accuracy with the same computational complexity, second, efficient implementation of the filter&#x...

This thesis addresses some generalizations of the nonlinear filtering problem that appear in real world scenarios when tracking moving objects. In particular, the generalizations include admittance of the object being undetected, presence of extraneous measurements, or assumption that the object ...

This work focuses on measurement methods of compression strength during the treatment of venous and lymphatic diseases of lower extremities. The thesis provides an overview of currently used compression techniques and compares different approaches to pressure measurement, suitable sensors, and th...

The thesis topic is the analysis and testing of three chosen methods for the train position dynamic estimation with the usage of track database and sensors, that are usually a part of the standard train equipment. The result of these methods is matching of the measured&#x...