|
|
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.
|
|
|
Analysis of GPON frames using machine learning
Tomašov, Adrián ; Horváth, Tomáš (referee) ; Holík, Martin (advisor)
Táto práca sa zameriava na analýzu vybraných častí GPON rámca pomocou algoritmov strojového učenia implementovaných pomocou knižnice TensorFlow. Vzhľadom na to, že GPON protokol je definovaný ako sada odporúčaní, implementácia naprieč spoločnosťami sa môže líšiť od navrhnutého protokolu. Preto analýza pomocou zásobníkového automatu nie je dostatočná. Hlavnou myšlienkou je vytvoriť systém modelov za použitia knižnice TensorFlow v Python3, ktoré sú schopné detekovať abnormality v komunikácií. Tieto modely používajú viaceré architektúry neuronových sietí (napr. LSTM, autoencoder) a zameriavajú sa na rôzne typy analýzy. Tento systém sa naučí na vzorovej vzorke dát a upozorní na nájdené odlišnosti v novozachytenej komunikácií. Výstupom systému odhad podobnosti aktuálnej komunikácie v porovnaní so vzorovou komunikáciou.
|
|
|
Autoencoder Implementation for Image Analysis
Sarančuk, Nikola ; Bilík, Šimon (referee) ; Horák, Karel (advisor)
The paper is devoted to the research of the problem of anomaly detection in industrial inspection. The paper describes the artificial neural network and its parts. The thesis contains a chapter where unary, binary and multi-class classifiers are compared. The thesis then explaines architecture of convolutional neural networks and autoencoder neural networks.. Then the paper describes the annotated dataset created. Finally, the paper describes the implementation of the convolutional autoencoder and evaluates the classification quality.
|
|
|
Non-Parallel Voice Conversion
Brukner, Jan ; Plchot, Oldřich (referee) ; Černocký, Jan (advisor)
Cílem konverze hlasu (voice conversion, VC) je převést hlas zdrojového řečníka na hlas cílového řečníka. Technika je populární je u vtipných internetových videí, ale má také řadu seriózních využití, jako je dabování audiovizuálního materiálu a anonymizace hlasu (například pro ochranu svědků). Vzhledem k tomu, že může sloužit pro spoofing systémů identifikace hlasu, je také důležitým nástrojem pro vývoj detektorů spoofingu a protiopatření. Modely VC byly dříve trénovány převážně na paralelních (tj. dva řečníci čtou stejný text) a na vysoce kvalitních audio materiálech. Cílem této práce bylo prozkoumat vývoj VC na neparalelních datech a na signálech nízké kvality, zejména z veřejně dostupné databáze VoxCeleb. Práce vychází z moderní architektury AutoVC definované Qianem et al. Je založena na neurálních autoenkodérech, jejichž cílem je oddělit informace o obsahu a řečníkovi do samostatných nízkodimenzionýálních vektorových reprezentací (embeddingů). Cílová řeč se potom získá nahrazením embeddingu zdrojového řečníka embeddingem cílového řečníka. Qianova architektura byla vylepšena pro zpracování audio nízké kvality experimentováním s různými embeddingy řečníků (d-vektory vs. x-vektory), zavedením klasifikátoru řečníka z obsahových embeddingů v adversariálním schématu trénování neuronových sítí a laděním velikosti obsahového embeddingu tak, že jsme definovali informační bottle-neck v příslušné neuronové síti. Definovali jsme také další adversariální architekturu, která porovnává původní obsahové embeddingy s embeddingy získanými ze zkonvertované řeči. Výsledky experimentů prokazují, že neparalelní VC na nekvalitních datech je skutečně možná. Výsledná audia nebyla tak kvalitní případě hi fi vstupů, ale výsledky ověření řečníků po spoofingu výsledným systémem jasně ukázaly posun hlasových charakteristik směrem k cílovým řečníkům.
|
|
|
Convolutional neural network for image processing
Krajčovičová, Mária ; Rajmic, Pavel (referee) ; Burget, Radim (advisor)
Goal of this Diploma thesis was Convolutional neural network investigation in last years. Diploma thesis also contains information about designing of appropriate Convolutional neural network models and implementation of these models in Java programming language. Result of the thesis are comparison and evaluation of results which were reached from implemented application.
|
|
|
Unsupervised Deep Learning Approach for Seizure Onset Zone localization in Epilepsy
Přidalová, Tereza ; Cimbálník, Jan (referee) ; Mehnen, Lars (advisor)
Epilepsy affects about 50 million people worldwide, with one-third of patients being drugresistant and therefore candidates for an invasive brain resection surgery. Brain resection surgery candidates undergo invasive intracranial encephalography (iEEG) monitoring to determine the seizure onset zone (SOZ). Recorded data can span over weeks and need to be manually reviewed by a physician to assess SOZ. This process can be time-consuming and burdensome due to the vast amount of collected data. This work investigates utilisation of an deep autoencoder for unsupervised data exploration and specifically its ability to discriminate between SOZ and non-SOZ (NSOZ) iEEG channels. The data used in this thesis consists of iEEG collected from 33 patients in two institutes (Mayo Clinic, Rochester, Minnesota, USA and St. Anne´s University Hospital, Brno, Czech Republic - FNUSA) who underwent invasive presurgical monitoring. The autoencoder’s capability to discriminate between SOZ and NSOZ was evaluated using a self-learned embedded feature space representation of the autoencoder network. Autoencoder features were compared to previously established biomarkers for SOZ determination. Discrimination capability was evaluated for both autoencoder features and biomarkers using a Naive Bayes classifier and leave-one-out cross-validation. The achieved area under receiver operating characteristic curve (AUROC) was 0.68 for the FNUSA and 0.56 for the Mayo dataset. Performance in discriminating between SOZ and NSOZ electrodes was not significantly different between the investigated autoencoder features and previously established biomarkers. Selecting the better performing classifier for each patient increased the AUROC to 0.75 and 0.64 for the FNUSA and Mayo dataset, respectively. The results suggest that future approaches combining biomarkers and self-learning methods have a potential to improve the SOZ vs NSOZ discrimination capability of unsupervised iEEG exploration systems, and thus to enhance the surgical management of epilepsy.
|
|
|
Time-Domain Neural Network Based Speaker Separation
Peška, Jiří ; Černocký, Jan (referee) ; Žmolíková, Kateřina (advisor)
A thesis is about the usage of convolutional neural networks for automatic speech separation in an acoustic environment. The goal is to implement the neural network by following a TasNet architecture in the PyTorch framework, train this network with various values of hyper-parameters, and to compare the quality of separations based on the size of the network. In contrast to older architectures that transformed an input mixture into a time-frequency representation, this architecture uses a convolutional autoencoder, which transforms input mixture into a non-negative representation optimized for a speaker extraction. Separation is achieved by applying the masks, which are estimated in the separation module. This module consists of stacked convolutional blocks with increasing dilation, which helps with modeling of the long-term time dependencies in processed speech. Evaluation of the precision of the network is measured by a signal to distortion (SDR) metric, by a perceptual evaluation of speech quality (PESQ), and the short-time objective intelligibility (STOI). The Wall Street Journal dataset (WSJ0) has been used for training and evaluation. Trained models with various values of hyper-parameters enable us to observe the dependency between the size of the network and SDR value. While smaller network after 60 epochs of training reached 10.8 dB of accuracy, a bigger network reached 12.71 dB.
|
|
|
Deep Neural Networks
Habrnál, Matěj ; Zbořil, František (referee) ; Zbořil, František (advisor)
The thesis addresses the topic of Deep Neural Networks, in particular the methods regar- ding the field of Deep Learning, which is used to initialize the weight and learning process s itself within Deep Neural Networks. The focus is also put to the basic theory of the classical Neural Networks, which is important to comprehensive understanding of the issue. The aim of this work is to determine the optimal set of optional parameters of the algori- thms on various complexity levels of image recognition tasks through experimenting with created application applying Deep Neural Networks. Furthermore, evaluation and analysis of the results and lessons learned from the experimentation with classical and Deep Neural Networks are integrated in the thesis.
|
|
|
Multiview Object Detection
Lohniský, Michal ; Beran, Vítězslav (referee) ; Juránek, Roman (advisor)
This thesis focuses on modification of feature extraction and multiview object detection learning process. We add new channels to detectors based on the "Aggregate channel features" framework. These new channels are created by filtering the picture by kernels from autoencoders followed by nonlinear function processing. Experiments show that these channels are effective in detection but they are also more computationally expensive. The thesis therefore discusses possibilities for improvements. Finally the thesis evaluates an artificial car dataset and discusses its small benefit on several detectors.
|
|
|
Unsupervised Anomaly Detection in Image
Salvet, Lukáš ; Herout, Adam (referee) ; Juránek, Roman (advisor)
This thesis deals with anomaly detection on industrial products. The main requirement was that the method required as little data with anomalies as possible at the time of construction and that it was easily applicable to different types of products. Neural network that is indirectly taught to find differences between two pictures is designed and described in this thesis. The anomaly detection itself should take place based on the representation of input data in latent space or in combination with a reconstruction loss. Four different method modifications have been designed and tested. The testing was mainly carried out on the MVTec AD dataset, which contains industrial products. Unfortunately the assumption that if the network is taught to look for differences the latent space will be interpreted better was not confirmed. Therefore the method was evaluated in a reconstructive error mode in~which it achieves comparable results with other methods. The result is insufficient for use in practice.
|
guest ::
