|
|
Detection of Harmfulness of Communication Partners and Their Networks
Kučera, Rostislav ; Homoliak, Ivan (referee) ; Očenášek, Pavel (advisor)
With the growing dependence of the population on electronic devices, the risk of data loss or misuse also increases. As the number of attacks in computer networks rises, systems for detecting malicious traffic become more important. The goal of this work is a theoretical analysis and implementation of modules for detecting malicious computer communication using machine learning methods, specifically a neural network model, and statistical analysis, which are deployed within the extended intrusion detection system Snort.
|
|
|
Use of neural networks for estimation of dynamic variables
Dufek, Martin ; Repka, Martin (referee) ; Zháňal, Lubor (advisor)
The aim of the thesis is to verify the applicability of neural networks to predict vehicle dynamic variables. Some vehicle dynamic variables are difficult to measure or need to be calculated, and measuring such quantities can be very expensive. However, If neural networks could predict values with acceptable error, this would be a more affordable and economical method. Verification was performed by creating two recurrent neural networks to estimate the quantities of directional deviation angle and longitudinal forces on all wheels of the car. The paper describes the steps of network creation from processing the input data to evaluating the network predictions. The results show that neural networks can be used to determine dynamic quantities and replace expensive measurements for some purposes. Finally, important insights gained during the creation of neural networks are formulated that can help with the creation of new networks for the estimation of automotive dynamic quantities, and further possible improvements of the created neural networks are outlined.
|
|
|
Semantic segmentation of aerial images
Pazdera, Jiří ; Králík, Jan (referee) ; Adámek, Roman (advisor)
This work deals with semantic segmentation of aerial images and their subsequent use for route planning. The first part represents an introduction to this issue and a theoretical description of the current state of knowledge. The second part describes testing of available segmentation methods, the development of custom dataset, and the training of an existing neural network model. Finally, the possibility of route planning using an appropriate algorithm is demonstrated.
|
|
|
Object Detection on the i.MX RT Microcontroller
Kravchuk, Marina ; Rozman, Jaroslav (referee) ; Janoušek, Vladimír (advisor)
This work focuses on the use of machine learning, particularly convolutional neural networks, in industrial applications. The course of work involves investigating the implementation of these networks directly on embedded devices, specifically NXP i.MX RT microcontrollers. During the course of the study, materials related to the training and use of neural networks and their optimization for deployment on low power devices were reviewed. Several neural network models were trained and tested, the best of which was used in the final version of the application. The application itself is divided into two parts: one part is written in C/C++ in the MCUXpresso IDE, where the main functionality of the program is implemented, while the other part of the work, i.e. the creation of a graphical user interface to control the program, is done in Python. The result is a functional application for the MIMXRT1170-EVK microcontroller that is able to detect and recognize small colored objects of certain shapes from a predefined data set.
|
|
|
Accelerometer data classification within the patient ECG record
Kindl, Zdeněk ; Ředina, Richard (referee) ; Bulková, Veronika (advisor)
The subject of the bachelor's thesis is the classification of patient accelerometric data. The aim is to improve the clarification of pathologies in the ECG signal. The classification is performed on data measured by the Bittium Faros 180L device. A custom database of movements was created. Patient data is processed using a recurrent neural network, which classifies the movements into three basic groups: resting activity, moderate activity, and high activity. The output is a file with movement annotations. The thesis includes a description of neural networks, data, data processing, and the creation of the neural network with codes.
|
|
|
Reconstruction of corrupted audio signals using deep unrolling
Kment, František ; Myška, Vojtěch (referee) ; Mokrý, Ondřej (advisor)
The thesis deals with the problem of audio signal restoration using traditional optimization methods combined with deep unrolling methods. An optimization task for filling in missing sections of the audio signal was formulated, and the proximal algorithm FISTA was chosen and subsequently implemented. Furthermore, three unrolled variants of the algorithm (Unrolled Fista Net) were implemented, two of which were further optimized using tests on a selected test dataset and trained on the Nsynth dataset. The results of the trained networks were compared both with competing methods and the original untrained variant of the algorithm. The comparison was made using objective metrics (MSE, SNR, PEAQ, PEMO-Q) and a subjective listening test.
|
|
|
Detection of Material Surface Damage Based on a Photograph
Marek, Radek ; Sakin, Martin (referee) ; Dyk, Tomáš (advisor)
This work focuses on the use of various types of neural networks for detecting surface damage of materials from photographs and evaluates their effectiveness. Identifying different types of damage, such as cracks, scratches, and other defects, is essential for assessing the condition of materials and may indicate the need for further maintenance or repairs. The use of advanced neural networks allows for more precise detection and classification of damage, which is crucial for applications in areas such as construction, the automotive industry, and aerospace engineering, where rapid and reliable diagnostics of material defects are critical. Integrating these technologies into regular inspection processes can significantly improve accident prevention and extend the lifespan of structural components. The work also discusses the possibilities for improvement and adaptation of algorithms to specific materials and types of damage. Thus, this work demonstrates how advanced machine learning technologies can significantly contribute to more effective and reliable material condition monitoring, opening paths for future innovations in maintenance and safety.
|
|
|
Potential of neural networks using transformers for medical image processing
Valík, Tomáš ; Nohel, Michal (referee) ; Chmelík, Jiří (advisor)
This thesis explores the potential of neural networks based on transformer architecture for medical image processing. The main objective was to compare the performance of ResNet18 and Vision Transformer (ViT-B-16) models on two distinct datasets, specifically Intel Image Classification and ChestXray. The models were optimized using the Optuna framework and subsequently trained ten times each to ensure robustness of the results. These results indicate that models utilizing Vision Transformers achieve higher weighted F1 scores compared to ResNet18 models. Specifically, the ViT-B-16 model achieved the highest F1 score of 0.939 on the Intel Image dataset and 0.907 on the ChestXray dataset, whereas ResNet18 achieved scores of 0.883 and 0.885, respectively. Statistical analyses using the Wilcoxon test confirmed that the differences in performance between the models are statistically significant, suggesting an advantage of using Vision Transformers for these tasks. An analysis of computational complexity is also provided, highlighting that ViT requires significantly higher computational resources.
|
|
|
Detekce začátku a konce komplexu QRS s využitím hlubokého učení
Müller, Jakub ; Šaclová, Lucie (referee) ; Smíšek, Radovan (advisor)
ECG measurement isan essential diagnostic tool for cardiac health, and automation of its analysis can aid to our healthcare to relieve staff workload or improve the quality of automated diagnostics from wearable devices. This work focuses specifically on the QRS complex in the ECG signal, with the main goal of using deep learning methods to detect its onset and offset. In the theoretical introduction, the reader is introduced to the origin of the QRS complex and ECG measurements, artificial neural networks and deep learning. Modified architecture U-Net for 1D signals was chosen to implement the actual method. Data were extracted from five publicly available databases and preprocessed in Matlab. This was followed by moving to the Python environment where parts of the model were implemented using the TensorFlow and Keras libraries, subsequent training, testing of the model and evaluation of the results.
|
|
| |
guest ::
