|
|
Generating training data with neural networks
Ševčík, Pavel ; Kolář, Martin (referee) ; Hradiš, Michal (advisor)
The aim of this thesis was to prepare a training data set for traffic sign detection using generative neural networks. The solution uses a modified U-Net architecture and several experiments with the application of styles using AdaIN layers as in the StyleGAN model have been conducted. By extending the real GTSDB data set with the generated images, mean average precision of 80.36 % has been achieved, which yields an improvement of 19.27 % compared to the mean average precision of the detection model trained on real data only.
|
|
|
Analysis of Human Signature Based on Artificial Neural Network
Ševčík, Pavel ; Horák, Karel (referee) ; Pohl, Jan (advisor)
This bachelor thesis deals with methods of human signature and its analysis in practical service of artificial neural network. Actual processing and analysis of human signature consist in few steps. First of all, the signature pattern is digitized and processed with the assistance of preprocessing and segmentation methods. Afterwards, the object of human signature pattern is described with the assistance of centric geometric moments and moments invariant characteristics. Finally, the pattern is classified by multilayer perceptron, whose outputs determine the person, to that signature belongs to.
|
|
|
Generative Neural Networks for Handwritten Text
Ševčík, Pavel ; Dobeš, Petr (referee) ; Hradiš, Michal (advisor)
The aim of this study was to create a generative neural network for handwritten text lines. The model produces variable-sized images of handwritten text lines based on the expected style. The proposed method exceeds existing models in the image quality and can be used to generate both individual words and entire lines of handwritten text. It combines the use of the attention mechanism to extract the features for each character from the text query and their arranging on the line by inserting spaces between them. The new approach allows more granular control of the symbol positions on the line, which leads to smoother style interpolations. In contrast to the previous approach, the proposed method uses the Gaussian filter to spread the individual symbols features to the surrounding area. This approach also allows to train the model for symbols position predictions using the adversarial loss (GAN). In addition, annotations of symbol horizontal positions on the lines of the IAM dataset of handwritten text have been created.
|
|
|
Design of the Information System for Operational Management
Ševčík, Pavel ; Krejčí, Jaromír (referee) ; Videcká, Zdeňka (advisor)
For the runtime phase of information system lifecycle is typical a frequent change of system quality requirements. Only systems that can adapt to these requirements are successful in its environment. The system change can be each time of different extensive. If it is necessary for satisfaction of new requirements to add, remove or change some system components and its relations than we talk about system architecture change. This action is so large-scale that it is very risky to implement it without appropriate control and management. This essay addresses these changes and its management. Main phases of the process and its activities are proposed. The suggestion is based on the development process of new systems and available literature from this area. The target of the essay is to offer a certain framework for the change process management.
|
|
|
Comfort and Safety
Ševčík, Pavel ; Mikulčík, Aleš (referee) ; Hájek, Vítězslav (advisor)
This bachelor’s thesis considers by safety and comfort systems in motor vehicles. Is concentrate to safety systems for example seat belts and tighteners, airbags and their separations and about direct units for activation safety systems. Next about system of active head rest and about safety transport of children in cars. Deal with comfort systems and shortly discuss about trend of safety systems. The last part of bachelor’s thesis considers of analyse and experimental measurement of dump sensor. Function of accelerometer ADXL150 is described and the practical measurement is made with including evaluation of measured results.
|
|
|
Single-Phase Induction Motor Calculation
Ševčík, Pavel ; Vítek, Ondřej (referee) ; Hájek, Vítězslav (advisor)
This master's thesis considers by single-phase asynchronous motor. The first part discuss about construction, principle of operation and basic parameters of this motor type. It also discussed the emergence of torque and torque characteristics of the different types of engines. The second part explains the basic ways of obtaining grip moment including outline the principles of functions particulars constructions types, which is produce in practice. In the third part is minutely analyse single-phase asynchronous motor with auxiliary phase and permanently conect capacitor, including method of assign size of a capacitor. Fourth part section provides the principles of the procedure for calculating basic parameters of the single asynchronous motor with auxiliary phase. At the last part is accomplished a calculation of single-phase asynchronous motor with permanently conect capacitor at auxiliary phase including comparation with already manufacturing motor.
|
|
|
Generative Neural Networks for Handwritten Text
Ševčík, Pavel ; Dobeš, Petr (referee) ; Hradiš, Michal (advisor)
The aim of this study was to create a generative neural network for handwritten text lines. The model produces variable-sized images of handwritten text lines based on the expected style. The proposed method exceeds existing models in the image quality and can be used to generate both individual words and entire lines of handwritten text. It combines the use of the attention mechanism to extract the features for each character from the text query and their arranging on the line by inserting spaces between them. The new approach allows more granular control of the symbol positions on the line, which leads to smoother style interpolations. In contrast to the previous approach, the proposed method uses the Gaussian filter to spread the individual symbols features to the surrounding area. This approach also allows to train the model for symbols position predictions using the adversarial loss (GAN). In addition, annotations of symbol horizontal positions on the lines of the IAM dataset of handwritten text have been created.
|
|
|
Generating training data with neural networks
Ševčík, Pavel ; Kolář, Martin (referee) ; Hradiš, Michal (advisor)
The aim of this thesis was to prepare a training data set for traffic sign detection using generative neural networks. The solution uses a modified U-Net architecture and several experiments with the application of styles using AdaIN layers as in the StyleGAN model have been conducted. By extending the real GTSDB data set with the generated images, mean average precision of 80.36 % has been achieved, which yields an improvement of 19.27 % compared to the mean average precision of the detection model trained on real data only.
|
|
|
Design of the Information System for Operational Management
Ševčík, Pavel ; Krejčí, Jaromír (referee) ; Videcká, Zdeňka (advisor)
For the runtime phase of information system lifecycle is typical a frequent change of system quality requirements. Only systems that can adapt to these requirements are successful in its environment. The system change can be each time of different extensive. If it is necessary for satisfaction of new requirements to add, remove or change some system components and its relations than we talk about system architecture change. This action is so large-scale that it is very risky to implement it without appropriate control and management. This essay addresses these changes and its management. Main phases of the process and its activities are proposed. The suggestion is based on the development process of new systems and available literature from this area. The target of the essay is to offer a certain framework for the change process management.
|
|
|
Analysis of Human Signature Based on Artificial Neural Network
Ševčík, Pavel ; Horák, Karel (referee) ; Pohl, Jan (advisor)
This bachelor thesis deals with methods of human signature and its analysis in practical service of artificial neural network. Actual processing and analysis of human signature consist in few steps. First of all, the signature pattern is digitized and processed with the assistance of preprocessing and segmentation methods. Afterwards, the object of human signature pattern is described with the assistance of centric geometric moments and moments invariant characteristics. Finally, the pattern is classified by multilayer perceptron, whose outputs determine the person, to that signature belongs to.
|
guest ::
RSS feed.