|
|
Robust Screen and Slide Detection in Video
Hanzel, Svätopluk ; Beran, Vítězslav (referee) ; Szőke, Igor (advisor)
The main goal of this bachelor thesis is implementation of a robust screen detector with slide synchronization using various techniques including neural networks, keypoints extraction and matching, text extraction using OCR and text matching. These methods are also analysed and compared to their possible alternatives.
|
|
|
Using machine learning for quality control in industrial applications
Gaško, Viktor ; Dobrovský, Ladislav (referee) ; Parák, Roman (advisor)
Goal of this bachelor´s thesis is to get acquainted with issue of quality control in industrial applications with focus on deep learning. For this and similar issues was created several libraries which have a purpose of simplifying these issues. Main task is to create program for quality control with help of programming language Python and framework Tensorflow. This program will be comprised of three neural network, from which one will identify the approximate position of the part, second its color, and third will check the correctness of its production.
|
|
|
Chatbot Based on Artificial Neural Networks
Richtarik, Lukáš ; Beneš, Karel (referee) ; Szőke, Igor (advisor)
This work deals with the issue of chatbots, which are based on artificial neural networks and generative models. It also describes options and process of designing the chatbot as well as an implementation and testing using BLEU metrics. The work contains multiple experiments with different models of chatbots, their performance evaluation and comparison, user experience and several suggestions for future enhancements.
|
|
|
Automatic classification of pronunciation of the letter „R“
Hrušovský, Enrik ; Vičar, Tomáš (referee) ; Harabiš, Vratislav (advisor)
This diploma thesis deals with automatic clasification of vowel R. Purpose of this thesis is to made program for detection of pronounciation of speech defects at vowel R in children. In thesis are processed parts as speech creation, speech therapy, dyslalia and subsequently speech signal processing and analysis methods. In the last part is designed software for automatic detection of pronounciation of vowel R. For recognition of pronounciation is used algorithm MFCC for extracting features. This features are subsequently classified by neural network to the group of correct or incorrect pronounciation and is evaluated classification success.
|
|
|
Artificial Poet
Bančák, Michal ; Szőke, Igor (referee) ; Beneš, Karel (advisor)
The paper presents a work on automatic poetry generation using the Long Short-TermMemory recurrent neural network. The aim of this work is to create an application thatimitates the writing of poems. This is a character-level language modeling in the Slovaklanguage. The neural network model used in the work consists of three layers of LSTM,with 400 hidden units. A collection of poems in the Slovak language with a size of 900k characters was also created for this work. . The final model is generating text that has poemelements. Achieved accuracy of generation is 41.85%.
|
|
|
Automation of Verification Using Artificial Neural Networks
Fajčík, Martin ; Husár, Adam (referee) ; Zachariášová, Marcela (advisor)
The goal of this thesis is to analyze and to find solutions of optimization problems derived from automation of functional verification of hardware using artificial neural networks. Verification of any integrated circuit (so called Design Under Verification, DUV) using technique called coverage-driven verification and universal verification methodology (UVM) is carried out by sending stimuli inputs into DUV. The verification environment continuously monitors percentual coverage of DUV functionality given by the specification. In current context, coverage stands for measurable property of DUV, like count of verified arithemtic operations or count of executed lines of code. Based on the final coverage, it is possible to determine whether the coverage of DUV is high enough to declare DUV as verified. Otherwise, the input stimuli set needs to change in order to achieve higher coverage. Current trend is to generate this set by technique called constrained-random stimulus generation. We will practice this technique by using pseudorandom program generator (PNG). In this paper, we propose multiple solutions for following two optimization problems. First problem is ongoing modification of PNG constraints in such a way that the DUV can be verified by generated stimuli as quickly as possible. Second one is the problem of seeking the smallest set of stimuli such that this set verifies DUV. The qualities of the proposed solutions are verified on 32-bit application-specific instruction set processors (ASIPs) called Codasip uRISC and Codix Cobalt.
|
|
|
Image classification using artificial intelligence
Labuda, Adam ; Přinosil, Jiří (referee) ; Burget, Radim (advisor)
This bachelor's thesis address the issue of classification and feature extraction of imagesfrom image. In JAVA platform will create an example that loads a set of images, extracted from symptoms with the help of artificial intelligence provided by the thesis supervisor. Artificial intellihence assumed kind of image. Finally the results are compared. }
|
|
|
GPGPU parallel computing
Pacura, Dávid ; Horák, Karel (referee) ; Petyovský, Petr (advisor)
The aim of this trim’s thesis is to reveal possibilities and demonstrate parallelization of computation on graphics processors. The paper presents descriptions of available development tools, and then one of them is selected to implement MD5 encryption algorithm and neural network for optical character recognition. Its performance is then compared to its parallel equivalent for conventional processors. In conclusion, problems encountered during development are described, and ways of avoiding them are discussed.
|
|
|
Neural Network Letter Recognition
Kluknavský, František ; Hradiš, Michal (referee) ; Šilhavá, Jana (advisor)
This work uses handwritten character recognition as a model problem for using multilayer perceptron, error backpropagation learning algorithm and finding their optimal parameters, hidden layer size, learning rate and length, ability to handle damaged data. Results were acquired by repeated simulation and testing the neural network using 52,152 English lowercase letters. Best results, smallest network and shortest learning time was at 60 neurons in the hidden layer and learning rate of 0.01. Bigger networks achieved the same ability to recognize unknown patterns and higher robustness at highly damaged data processing.
|
|
| |
guest ::
