|
|
Playing Games Using Neural Networks
Buchal, Petr ; Kolář, Martin (referee) ; Hradiš, Michal (advisor)
The aim of this bachelor thesis is to teach a neural network solving classic control theory problems and playing the turn-based game 2048 and several Atari games. It is about the process of the reinforcement learning. I used the Deep Q-learning reinforcement learning algorithm which uses a neural networks. In order to improve a learning efficiency, I enriched the algorithm with several improvements. The enhancements include the addition of a target network, DDQN, dueling neural network architecture and priority experience replay memory. The experiments with classic control theory problems found out that the learning efficiency is most increased by adding a target network. In the game environments, the Deep Q-learning has achieved several times better results than a random player. The results and their analysis can be used for an insight to reinforcement learning algorithms using neural networks and to improve the used techniques.
|
|
| |
|
|
Search for Duplicities and Quality Evaluvation of Photos
Sklenář, Zdeněk ; Hradiš, Michal (referee) ; Zemčík, Pavel (advisor)
This bachelor thesis is about the analysis, design, and implementation and testing of an application, which is used to find duplicates in photographs according to it's Exif metadata. The app can also let you preview the photo, including Exif metadatas. It is possible to filter your photos. You can group duplicates with the original photo, and select the best photos to keep it in accordance with a user-defined parameter, then manually adjust this option, and deleting others. There is also a possibility to export selected photos to a ZIP archive.
|
|
|
Board Game User Interface with a Camera
Cihlářová, Dita ; Hradiš, Michal (referee) ; Zemčík, Pavel (advisor)
Cílem této práce je vytvořit systém, který bude schopen nahradit reálného protihráče ve hře Dostihy a sázky. Hra je snímána kamerou, ze které se obraz přenáší do počítače, kde probíhá zpracování obrazu a identifikace herních objektů. Dále aplikace analyzuje stav hry a rozhodne o svém následujícím tahu. Rozhodovací algoritmus je schopen reagovat na každou herní situaci a tudíž dohrát hru do konce. Aplikaci mohou využít hráči, jež rádi hrají Dostihy a sázky, ale chybí jim protihráč. Aplikace také může najít uplatnění jako demostrace možností počítačového vidění.
|
|
|
Facial image restoration
Bako, Matúš ; Herout, Adam (referee) ; Hradiš, Michal (advisor)
In this thesis, I tackle the problem of facial image super-resolution using convolutional neural networks with focus on preserving identity. I propose a method consisting of DPNet architecture and training algorithm based on state-of-the-art super-resolution solutions. The model of DPNet architecture is trained on Flickr-Faces-HQ dataset, where I achieve SSIM value 0.856 while expanding the image to four times the size. Residual channel attention network, which is one of the best and latest architectures, achieves SSIM value 0.858. While training models using adversarial loss, I encountered problems with artifacts. I experiment with various methods trying to remove appearing artefacts, which weren't successful so far. To compare quality assessment with human perception, I acquired image sequences sorted by percieved quality. Results show, that quality of proposed neural network trained using absolute loss approaches state-of-the-art methods.
|
|
|
Tool for Text Corrections
Zatloukal, Jakub ; Hradiš, Michal (referee) ; Zemčík, Pavel (advisor)
The goal of this thesis was to develop a simple application, which would allow inscription of proofreading marks, annotations and more information into an electronic document. The interface of this appliaction should be intuitive and simple enough to be efficiently used without studying complex manuals or necessity of training by other person. Next requirement was, that proofreading information should be easily shared electronically. Resulting application is able to save all the proofreading information into a file separately from a document and to insert prepared proofreading marks. The file containing proofsheet is smaller than whole document. The test proved that application is intuitive, simple and visually friendly, but less practically usable. Its main benefit is simple and straightforward interface, which is in a certain manner adapted to proofreader's needs.
|
|
|
Face Detection
Šašinka, Ondřej ; Hradiš, Michal (referee) ; Juránek, Roman (advisor)
This MSc Thesis deals with face detection in image. In this approach, facial features (eyes, nose, mouth corners) are detected first and then joined to the whole face. For the facial features detection, classifiers trained with AdaBoost algorithm are used. Haar wavelets are used as features for classification.
|
|
|
Building deep networks using autoencoders
Lohniský, Michal ; Veselý, Karel (referee) ; Hradiš, Michal (advisor)
This thesis deals with pretraining deep networks by autoencoders. Components of neural networks are described in first chapters. Rest of chapters aims to deep network trainings and to results of experiments where autoencoder pretraining and Backpropagation algorithm are compared. Results showed positive contribution of autoencoder pretraining, mainly in combination with Finetuning.
|
|
|
Traffic Signs Detection
Ťapuška, Tomáš ; Beran, Vítězslav (referee) ; Hradiš, Michal (advisor)
This bachelor's thesis is about traffic sign detection in picture. There are written some known methods, their advantages and disadvantages. There is present implementation of the system for traffic sign detection. There are present in the last chapter some tests that were done on the system with using testing set, which was created specialy for this purpose.
|
|
|
ASL Fingerspelling Recognition Using Slow Feature Analysis
Winkler, Martin ; Hradiš, Michal (referee) ; Burget, Lukáš (advisor)
Táto práca popisuje proces testovania slow feature analysis ako metódy, ktorá extrahuje robustné črty z komplexných obrazových dát americkej znakovej reči. Za účelom testovania bol vytvorený systém v programovacom jazyku python, ktorý zjednodušuje testovanie a ponúka bohatú škálu meniteľných parametrov aby umožnil užívateľovi rôzne testy za účelom zistenia nakoľko použiteľná je táto metóda na klasifikáciu a rozpoznávanie gest rúk. Teoretická časť predstaví slow feature analysis, diskutuje o štruktúre systému a popisuje dáta na ktorých bude metóda pozorovaná. V praktickej časti je metóda podrobená analýze úspešnosti na videných a nevidených rečníkoch, jej schopnosť adaptovať sa na vyšší počet gest a zaujímavé formátovanie dát v pokuse vylepšiť jej úspešnosť.
|
guest ::
