|
|
Computer Vision with Active Learning
Kolář, Martin ; Machová, Kristína (referee) ; Arridge, Simon (referee) ; Zemčík, Pavel (advisor)
Metody strojového vidění se zdokonalují zlepšením modelů, laděním trénovaných parametrů nebo anotací reprezentativních dat. Tato práce řadou experimentů potvrzuje hypotézu že aktivní učení zvyšuje přesnost těchto modelů. Rozšířením přistupu pseudolabelů o aktivní učení přispívá tato práce přístupem "one-shot-learning" k učení nových kategorií obrazů s použitím algoritmických doporučení, dále online grafickým uživatelským rozhraním pro optimalizaci dilema Exploration/Exploitation pro online tagování, a dvoukrokovým offline binárním přístupem aktivního učení pro zlepšení kvality dat používaných pro snímání fontů. Tím, že demonstruje přínos aktivního učení v těchto přístupech, přispívá tato práce k hypotéze i konkrétním aplikacím strojového vidění.
|
|
|
Mobile Application for Gyms
Hynek, Tomáš ; Kolář, Martin (referee) ; Kapinus, Michal (advisor)
The aim of this bachelor’s thesis is to develop a mobile application for Android that allows you to build a training plan and record completed exercises. This thesis is aimed at beginners or advanced exercisers from Czech Republic. The result is a mobile app that motivates exercisers to exercise regularly and record their activities. Among other things, app allows the user to view more than 100 exercises accompanied by images and procedure description.
|
|
|
Network Communications for Virtual Waiting Room
Novák, Lukáš ; Kolář, Martin (referee) ; Brejcha, Jan (advisor)
The core of this thesis is to describe in detail the design of communication protocol for client/server communication and creation of a functioning server application prototype. REST design approach was used for this communication. This approach uses mainly GET and POST functions of HTTP transport protocol. The server application prototype is created in PHP with the help of Silex micro-framework. The protocol is designed for mobile clients, which are used for registering into queues and also for administrative clients, which can control and manage the queues. Testing of the designed protocol with server application prototype is described at the end of this thesis.
|
|
|
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This thesis deals with text line recognition of historical documents.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.
|
|
|
Birds Eye Panorama
Sobotka, Lukáš ; Pavelková, Alena (referee) ; Kolář, Martin (advisor)
The goal of this paper is to create panorama from the collection of photographies, which are taken from randomly moving device with OS Android. The device is placed on flying vehicle (a kite, a helium balloon, a drone etc.). The photographies are taken from birds eye view. The Android application was created and tested. This application creates a collection of photographies and then runs an algorithm to create panorama. This algorithm specifies descriptors using SURF, which are filtered by RANSAC algorithm for finding the optimal homography.This algorithm is perfomed iteratively over the collection. The image data is processed using the OpenCV library.
|
|
|
Application for Tree Stability Measurement
Šoltés, Jakub ; Kolář, Martin (referee) ; Polok, Lukáš (advisor)
This thesis is concerned with the subject of tree stability under a wind load. It defines, what the tree stability is, which techniques are used for its measurement and why it is important to determine it. It describes development process of application, that examines tree stability. Furthermore it analyses available the development tools, discusses a design and an implementation of the application written in C# language. Application has been created using .NET Framework along with graphical subsystem Windows Presentation Foundation following the design pattern Model View ViewModel. The result of this thesis is a desktop application aimed on Windows operating system. It can compute stability coefficient of the tree using information about trees‘ geometry and its environment. The value of stability coefficient is commonly used for tree safety inspection.
|
|
|
Virtual Waiting Room Client for Service Provider
Holcman, Jaroslav ; Brejcha, Jan (referee) ; Kolář, Martin (advisor)
The goal of this thesis is to develop and implement client application with graphical user interface for service providers as a part of the project Virtual waiting room. Graphical user interface enables the provider to manage queues of costumers, appoint costumers for specific terms and work with data stored on the server. The user interface is developed using the WPF technology from the .NET Framework, created by Microsoft. This thesis describes the Waiting room project, the client for service provider and also technologies used to develop this client.
|
|
|
Generating Faces with Generative Adversarial Networks
Konečný, Daniel ; Herout, Adam (referee) ; Kolář, Martin (advisor)
Cílem této práce je generování barevných obrázků obličejů z náhodně určených vysokodimenzionálních vektorů pomocí generativních neuronových sítí. Dále se zabývá analýzou vstupních vektorů na základě příznaků obličejů z nich vygenerovaných. Je provedena implementace generativní neuronové sítě pro generování obrázků ručně psaných číslic, poté dalších dvou sítí pro generování obrázků obličejů. Vygenerované obrázky zobrazují věrohodně vypadající obličeje, lidské oko je však dokáže odlišit od fotek reálných osob. Analýza jednotlivých dimenzí vektorů je provedena pomocí Studentova t-testu. Dále jsou vstupní vektory promítnuty do podprostorů pomocí lineární diskriminační analýzy a jsou nalezeny rozdělovací hranice mezi třídami příznaků. Analýza generovaných dat dokazuje, že ovlivněním vstupního vektoru je možné docílit generování obrázku obličeje s požadovanými příznaky s pravděpodobností až 80 %. Hlavním výsledkem této práce je model generativní neuronové sítě určené pro generování obrázků obličejů. Dalším přínosem je nástroj pro generování obrázků obličejů na základě vybraných příznaků.
|
|
|
Personal Calendar PHP
Sadloň, Matej ; Kapinus, Michal (referee) ; Kolář, Martin (advisor)
This thesis aims to time management. The goal is to design and implement web application for simpler and easy-to-use managing of personal time. For this, the application uses calendars, To-Do lists, dividing of bigger task into subtasks or analyzing of spent time. The PHP framework named Symfony was used for implementation of the application with use of object-relation mapping, that is provided by the PHP library Doctrine and uses MySQL database. The CSS and JavaScript technologies were used for better graphical design and dynamism of work with application.
|
|
|
Altitude Estimation from an Image
Vašíček, Jan ; Kolář, Martin (referee) ; Čadík, Martin (advisor)
This thesis is concerned with the automatic altitude estimation from a single landscape photograph. I solved this task using convolutional neural networks. There was no suitable training dataset available having information about image altitude, thus I had to create a new one. To estimate human performance in altitude estimation task, an experiment was conducted counting 100 subjects. The goal of this experiment was to measure the accuracy of the human estimate of camera altitude from an image. The measured average estimation error of subjects was 879 m. An automatic system based on convolutional neural networks outperforms humans with an average elevation error 712 m. The proposed system can be used in more complex scenario like the visual camera geo-localization.
|
guest ::
RSS feed.