|
|
Recognition and Filtration of Unwanted Video-Sequences
Vlček, Jakub ; Surynek, Pavel (advisor) ; Kolomazník, Jan (referee)
The thesis deals with the possibilities of recognition video sequences and possibility of machine learning based on previous knowledge. I use neural networks which I learn on characteristics generated from images. Images are characterized using image functions like histogram and statistical functions such as mean and correlations to detect scenes in video-sequence. I used all of these findings to implement the demonstration program. The program can play video, display statistical information about the sequence or frame by frame and detect sequences similar to advertisements on the basis of statistical properties and respond to it by defined events.
|
|
| |
|
| |
|
|
Dynamic simulation of rigid bodies using programmable GPUs
Cséfalvay, Szabolcs ; Kmoch, Petr (advisor) ; Kolomazník, Jan (referee)
The goal of this work is to create a program which simulates the dynamics of rigid bodies and their systems using GPGPU with an emphasis on speed and stability. The result is a physics engine that uses the CUDA architecture. It runs entirely on the GPU, handles collision detection, collision response and different forces like friction, gravity, contact forces, etc. It supports spheres, rods (which are similar to cylinders), springs, boxes and planes. It's also possible to construct compound objects by connecting basic primitives.
|
|
|
Vessel segmentation
Dupej, Ján ; Pelikán, Josef (advisor) ; Kolomazník, Jan (referee)
Title: Vessel segmentation Author: Ján Dupej Department / Institute: Department of Software and Computer Science Education Supervisor of the master thesis: RNDr. Josef Pelikán, KSVI Abstract: In this thesis we researched some of the blood vessed segmentation and visualization techniques currently available for angiography on CT data. We then designed, implemented and tested a system that allows both semi-automatic and automatic vessel segmentation and visualization. For vessel segmantation and tracking we used a region-growing algorithm that we overhauled with several heuristics and combined with centerline detection. We then automated this algorithm by automatic seed generation. The visualization part is accomplished with an adaptation of the well-known straightened CPR method that we enhanced so that it visualizes the whole cross-section of the blood vessel, instead of just one line of it. Furthermore, we used the Bishop frame to maintain minimal twist of the curve-local coordinate system along the whole vessel. Keywords: vessel segmentation, medical data analysis, volume data
|
|
|
Face Recognition in Social Networks
Mach, Pavel ; Skopal, Tomáš (advisor) ; Kolomazník, Jan (referee)
Popularity of face recognition in image galleries and social networks is growing. But none of these real world applications let the user adjust parameters. The goal of this thesis is to develop a library for face detection and recognition which can be easily used in some other program and also a web application recognizing faces, with environment similar to the one of social networks, which uses the library. Of course we do not try to compete with big social networks in the field of face recognition, because it will be almost impossible due to their enormous databases. In the experimental part of the work we try to find most suitable parameters such that face detection and face recognition would be as accurate as possible.
|
|
|
Generování vlasů interpolací
Šik, Martin ; Křivánek, Jaroslav (advisor) ; Kolomazník, Jan (referee)
This thesis describes a procedural hair generator that is able to generate hair from just a few hairs, called hair guides, which are directly modeled by a 3d artist. The procedural hair generator is a part of Stubble project -- a tool for hair modeling in Autodesk Maya. The procedural hair generator can generate hair during rendering, thus avoiding storage of hair geometry in a scene file, which makes the rendering process very efficient. Furthermore, hair can be generated interactively and displayed by OpenGL during modeling in Maya. Generated hair geometry is mainly defined by interpolation from the mentioned hair guides; however it is also influenced by many hair properties. These properties can change hair geometry using noise functions, define hair color, width and more. To determine hair root positions on a given triangular mesh I use my own mesh sampling algorithm that generates random samples on a triangular mesh according to a density defined by a 2-dimensional texture. My sampling algorithm uses an innovative way of sampling from a discrete probability distribution, which can be used in other applications than mesh sampling.
|
|
|
Rigid body simulation
Zvirinský, Peter ; Pelikán, Josef (advisor) ; Kolomazník, Jan (referee)
The object of this work is to create an easily extendible rigid body simulation engine. The engine will contain real-time collision detection and rea- listic collision handling. It will be able to simulate common nature forces such as gravity, related different types of friction and resting forces. The simulation engine will mainly focus on spheres and their movement influenced by the forces mentioned before. The part of this work will be as well the examination of diffe- rent types of optimalization of such tasks as collision detection and use of parallel approach. 1
|
|
|
Data extraction from document scans
Macháč, Bohuslav ; Kolomazník, Jan (advisor) ; Krajíček, Václav (referee)
In this work I developed an application capable of extracting data from scanned documents. For optical character recognition, I used external OCR engine Tesseract, but it can be easily changed. I use document templates, which have informations about data areas and its data types. I tried to automatize most of the steps which are required to extract data or create new data template. User can improve or change results of these steps. For export from application I implemented components, which export data to XML, HTML or plain text. Another components can be easily added, to adapt application for various uses.
|
|
|
Character recognition of machine-written documents
Kindermann, Hubert ; Blažek, Jan (advisor) ; Kolomazník, Jan (referee)
In the present thesis we solve the problem of symbol extraction and recognition from printed documents digitized by the scanner or camera. We introduce a noise resistant algorithm of document lighting normalization. We continue with the extraction of individual characters from the document and their recognition with a system of feedforward multilayer neural networks. We also focus on processing of the resulting set of recognized characters, which is necessary for further use of the extracted text. The last step is correction of the output based on surrounding letters of each character. We have successfully implemented an automatic system containing all the above components.
|
guest ::
RSS feed.