|
Convolutional Networks for Historic Text Recognition
Vešelíny, Peter ; Kolář, Martin (referee) ; Kišš, Martin (advisor)
This thesis deals with text line recognition of historical documents. Historical texts dating back to the 17th - 19th centuries are written in fraktur typeface. The character recognition problem is solved using neural network architecture called sequence-to-sequence . This architecture is based on encoder-decoder model and contains attention mechanism. In this thesis a dataset, from texts originated from German archiv called Deutsches Textarchiv , was created. This archive contains 3 897 different German books that have available transcripts and corresponding images of pages. The created dataset was used to train and experiment with the proposed neural network. During the experiments, several convolutional models, hyperparameters and the effects of positional embedding were investigated. The final tool can recognize characters with accuracy 99,63 %. The contribution of this work is the~mentioned dataset and neural network, which can be used to recognize historical documents.
|
|
Generative Adversial Network for Artificial ECG Generation
Šagát, Martin ; Ronzhina, Marina (referee) ; Hejč, Jakub (advisor)
The work deals with the generation of ECG signals using generative adversarial networks (GAN). It examines in detail the basics of artificial neural networks and the principles of their operation. It theoretically describes the use and operation and the most common types of failures of generative adversarial networks. In this work, a general procedure of signal preprocessing suitable for GAN training was derived, which was used to compile a database. In this work, a total of 3 different GAN models were designed and implemented. The results of the models were visually displayed and analyzed in detail. Finally, the work comments on the achieved results and suggests further research direction of methods dealing with the generation of ECG signals.
|
|
Detection of Vehicle License Plates in Video
Líbal, Tomáš ; Hradiš, Michal (referee) ; Herout, Adam (advisor)
This thesis deals with preparation of training dataset and training of convolutional neural network for licence plate detection in video. Darknet technology was used for detection, specifically the YOLOv3-tiny neural network model. The solution was focused on the most accurate detection and the smallest number of false positives per image, thus minimizing overall model error. Dataset was prepared from existing freely available datasets, from the dataset provided by the GRAPH@FIT research group, and from self-annotated images created from downloaded YouTube videos. Furthermore, this dataset has been processed using data augmentation, extending it to twice the size. The YOLO Mark tool was used to create annotations. An ROC curve was used to visualize the detection success. Created solution reaches minimum total error 10,849%. Part of the solution is already mentioned dataset.
|
|
Image segmentation using deeplearning methods
Lukačovič, Martin ; Burget, Radim (referee) ; Mašek, Jan (advisor)
This thesis deals with the current methods of semantic segmentation using deep learning. Other approaches of neaural networks in the area of deep learning are also discussed. It contains historical solutions of neural networks, their development, and basic principle. Convolutional neural networks are nowadays the most preferable networks in solving tasks as detection, classification, and image segmentation. The functionality was verified on a freely available environment based on conditional random fields as recurrent neural networks and compered with the deep convolutional neural networks using conditional random fields as postprocess. The latter mentioned method has become the basis for training of new models on two different datasets. There are various enviroments used to implement neural networks using deep learning, which offer diverse perform possibilities. For demonstration purposes a Python application leveraging the BVLC\,/\,Caffe framework was created. The best achieved accuracy of a trained model for clothing segmentation is 50,74\,\% and 68,52\,\% for segmentation of VOC objects. The application aims to allow interaction with image segmentation based on trained models.
|
|
Deep Neural Network Pruning for Text Recognition
Petráš, Simon ; Hradiš, Michal (referee) ; Kišš, Martin (advisor)
This document is a work on pruning neural network for handwriting recognition. The aim of the work is to create a program for pruning the network. We prune two types of neural networks, namely convolutional and recurrent neural networks. During the pruning of the convolution part, various criteria of parameter selection were experimented with. The result of the work is a model that achieves 20% acceleration while increasing the network inaccuracy by only 0.4%, but also a number of other models that are faster but also acquire higher inaccuracies.
|
|
Automatická adaptace pouličního osvětlení na základě dat z kamer
Švanda, Jan
The thesis deals with the creation of a system that, based on the analysis of the real situation on the road using image data from cameras can provide data for the prediction of the optimal lighting settings for each moment. The thesis also deals with the selection and comparison of detection methods and subsequent optimization of the selected method.
|
|
The Russian invasion of Ukraine through the eyes of selected media outlets and theirs accounts on social media platforms
Helclová, Anna ; Géla, František (advisor) ; Lokšík, Martin (referee)
The central theme of the thesis is the coverage of the beginning of the invasion of Ukraine, the liberation of the city of Buch and the first foreign trip of Ukrainian President Volodymyr Zelensky by BBC News, CNN and CT24 and their social media accounts on Facebook and Instagram. The theoretical part of the thesis introduces the concepts of war and peace journalism, the transformation of social networks into a new source of information and the related information overload and the role of social networks in its formation. It also explains the circumstances of selected events for further analysis in the practical part of the thesis. In the research part, codes are defined according to which the posts of the selected media outlets are subsequently analyzed in the specified time windows on Instagram and Facebook. The posts are examined both with respect to their graphical form and their content. For video posts, the origin of the video is distinguished, as well as its content and its' most prominent features. Emphasis is placed on posts with drastic content and each media outlet's approach to such material. The analysis also takes into account the frequency of posts published by each media in selected time windows. The paper concludes with the results of the analysis of the posts and their interpretation.
|
|
Comparative analysis of differences between western and eastern news outlets reporting on the Russian invasion of Ukraine
Neumann, Jakub ; Koblovský, Petr (advisor) ; Shavit, Anna (referee)
This bachelor thesis focuses on the analysis of the rhetoric of the Izvestia media in response to the Russian invasion of Ukraine. Based on a careful examination and critical analysis of the articles examined, this bachelor's thesis compares and contrasts the ways in which Western and Eastern media outlets have dealt with this conflict, especially in light of their news interests and propaganda. It then goes on to map which narratives the Izvestia media use and which are most common in their articles. The results show the frequent use of false or misleading claims by the news outlet Izvestia. They also show the different portrayal of reality between Western and Eastern media in specific analyzed articles.
|
|
Perception of the "culture wars" concept in American media
Šimek, Jan ; Just, Petr (advisor) ; Charvát, Jakub (referee)
American society is going through an era of deep social and political divisions. At the forefront of this deepening divide are the so-called culture wars, an umbrella term for a set of narratives around cultural and social issues that both sides of the conflict use to advance their political goals. This thesis aims to determine how much space culture wars themes occupy in the production of selected American media outlets, how these themes are projected onto their audiences and how this influences the audience's trust in the mainstream media. The first part of the thesis presents the methodology used in collecting and evaluating the data used in the research for this thesis. The theoretical section also defines concepts and terms related to the topic of the thesis. The practical part of the thesis will then focus on the evaluation of the data from the questionnaire survey conducted among American media consumers between May and June 2023.
|
|
Automatic diagnosis of the 12-lead ECG using deep learning
Blaude, Ondřej ; Chmelík, Jiří (referee) ; Provazník, Valentine (advisor)
The aim of this diploma thesis is to investigate the problematics of automatic ECG diagnostics, namely on twelve-lead recordings. This problem is solved by standard methods such as random forest, artificial neural networks or K-nearest neighbors. However, thanks to its ability to independently extract symptoms, deep learning methods are also popular. All these methods are described in the theoretical part. In the practical part, deep learning models were designed, functionality support was verified using data from the PhysioNet database. Two pilot models were created and subsequently optimized. From the entire parameter optimization procedure, three models are available, of which the best accuracy achieves an F1 score of 87.35% and 83.7%, and the second best achieves an F1 score of 77.74% and an accuracy of 84.53%. The results achieved are discussed and compared with those of similar publications.
|