|
Neural Network Based Image Segmentation
Vrábelová, Pavla ; Žák, Pavel (referee) ; Švub, Miroslav (advisor)
This paper deals with application of neural networks in image segmentation. First part is an introduction to image processing and neural networks, second part describes an implementation of segmentation system and presents results of experiments. The segmentation system enables to use different types of classifiers, various image features extraction and also to evaluate the success of segmentation. Two classifiers were created - a neural network (self-organizing map) and an algorithm K-means. Colour (RGB and HSV) and texture features and their combinations were used for classification. Texture features were extracted using a set of Gabor filters. Experiments with designed classifiers and feature extractors were carried out and results were compared.
|
|
Neural Networks and Their Applications
Chaloupka, David ; Rozman, Jaroslav (referee) ; Zbořil, František (advisor)
The aim of this thesis is to present a consistent insight into the most frequently used types of artificial neural networks and their applications. It depicts feedforward neural networks with backpropagation training algorithm, Hopfield networks and self-organizing maps (Kohonen maps). Second part of this thesis demonstrates typical applications of described networks and discusses various factors, which influence performance of these networks on chosen tasks.
|
|
Criminality Analysis in the Czech Republic using Self-Organizing Maps
Mikulíková, Pavla ; Cahlík, Tomáš (advisor) ; Palanská, Tereza (referee)
Crime represents one of the most persistent social problems all around the world. To understand the motivation for criminal behaviour, a thorough analysis of its plausible determinants is necessary. This bachelor thesis aims at exploring whether the method of self-organizing maps, a data mining tool, can help in the investigation of the Czech criminal phenomena. To date, no academic study has tried to uncover potential pat- terns in the Czech crime data employing this type of artificial neural network. It is a visualisation method which maps observations based on their multi-dimensional features into a two-dimensional grid, and at the same time, the similarity between observations is preserved by locating similar observations close to each other. For the analysis, the dataset consisting of 75 Czech districts and 18 variables was used. However, the optimal choice of parameters of the model can be seen as a possible limitation of this method. The final outcome of the model consists of six clusters of districts with various levels of crime rates and other characteristics. Our results showed that self-organizing maps can provide an interesting insight into the crime problem, and social sciences can benefit from its application in many research areas. 1
|
|
Neural Networks and Their Applications
Chaloupka, David ; Rozman, Jaroslav (referee) ; Zbořil, František (advisor)
The aim of this thesis is to present a consistent insight into the most frequently used types of artificial neural networks and their applications. It depicts feedforward neural networks with backpropagation training algorithm, Hopfield networks and self-organizing maps (Kohonen maps). Second part of this thesis demonstrates typical applications of described networks and discusses various factors, which influence performance of these networks on chosen tasks.
|
|
Neural Network Based Image Segmentation
Vrábelová, Pavla ; Žák, Pavel (referee) ; Švub, Miroslav (advisor)
This paper deals with application of neural networks in image segmentation. First part is an introduction to image processing and neural networks, second part describes an implementation of segmentation system and presents results of experiments. The segmentation system enables to use different types of classifiers, various image features extraction and also to evaluate the success of segmentation. Two classifiers were created - a neural network (self-organizing map) and an algorithm K-means. Colour (RGB and HSV) and texture features and their combinations were used for classification. Texture features were extracted using a set of Gabor filters. Experiments with designed classifiers and feature extractors were carried out and results were compared.
|
| |
|
Cluster analysis of more dimensional data by a neural network
Helcl, Zbyněk ; Křivan, Miloš (advisor) ; Berka, Petr (referee)
The topic of the present thesis is an analysis of a sample data archive containing measured values of real and reactive power. The measurement in question took place in late 2006 and early 2007 using MEg40 recording measurement devices disposed in a station for transforming high voltage to low voltage in the Pražská energetika distribution network. The procedure of processing measured values, the preparation thereof for a subsequent processing by a neural network, and a final statistical evaluation of determined individual clusters -- typical daily take-off diagrams -- will be described. The results of the present thesis may be applied in the making of predictions of electrical energy consumption at a particular transformer station.
|