|
|
Vehicle Control via Reinforcement Learning
Maslowski, Petr ; Uhlíř, Václav (referee) ; Šůstek, Martin (advisor)
The goal of this thesis is a creation of an autonomous agent that can control a vehicle. The agent utilizes reinforcement learning that uses neural networks. The agent interprets images from the front vehicle camera and selects appropriate actions to control the vehicle. I designed and created reward functions and then experimented with hyperparameters setup. Trained agent simulate driving on the road. The result of this thesis shows a possible approach to control an autonomous vehicle agent using machine learning method in CARLA simulator.
|
|
|
Mobile Application for Recommending and Managing Cooking Recipes
Lončík, Andrej ; Šůstek, Martin (referee) ; Zbořil, František (advisor)
The goal of the submitted thesis is the creation of mobile application for devices using the Android operation system. The main purpose of the application is the discovery and administration of food recipes and meal planning. The functions of the application include voice control and search by a photo or an image. This work describes the whole process of app -creation, beginning from the original idea, followed by the competition analysis, draft of the user interface, its implementation and concluding with the testing and final publication to the Google Play. In addition, the final version of the application offers the feature of creating new recipes or searching for already published ones on the internet based on the ingredients the user possesses. The ingredients can be written in, entered by the user's voice, or recognized from an uploaded image. The photo and image recognition is provided by the Firebase ML Kit Image Labeling tool. Thanks to the Google account authentization , the application is also able to save the user's content in Firebase Realtime Database. Mobile application is published on the Google Play store and is officially named Recipio .
|
|
|
Movie Recommender System
Janko, Pavel ; Zbořil, František (referee) ; Šůstek, Martin (advisor)
This thesis primarily addresses various methods of constructing a system for movie recommendations. Both basic and advanced techniques required for creating a recommender system are also covered in the thesis. The core of the thesis is designing, implementing and experimenting with a system for movie recommendations based upon the data originating from publicly accessible datasets. In order to predict ratings that the user would give to movies after watching them, the system utilizes a factorization model based on collaborative filtering. This thesis also describes the relation between model hyperparameter configuration and prediction accuracy, experiments that were conducted in order to further improve the model accuracy and finally compares the implemented model with existing solutions.
|
|
|
Image Segmentation with Deep Neural Network
Pazderka, Radek ; Šůstek, Martin (referee) ; Rozman, Jaroslav (advisor)
This master's thesis is focused on segmentation of the scene from traffic environment. The solution to this problem is segmentation neural networks, which enables classification of every pixel in the image. In this thesis is created segmentation neural network, that has reached better results than present state-of-the-art architectures. This work is also focused on the segmentation of the top view of the road, as there are no freely available annotated datasets. For this purpose, there was created automatic tool for generation of synthetic datasets by using PC game Grand Theft Auto V. The work compares the networks, that have been trained solely on synthetic data and the networks that have been trained on both real and synthetic data. Experiments prove, that the synthetic data can be used for segmentation of the data from the real environment. There has been implemented a system, that enables work with segmentation neural networks.
|
|
|
Traveling Salesman Problem
Šůstek, Martin ; Snášelová, Petra (referee) ; Zbořil, František (advisor)
This thesis is focused on modification of known principles ACO and GA to increase their performance. Thesis includes two new principles to solve TSP. One of them can be used as an initial population generator. The appendix contains the implementation of the application in Java. The description of this application is also part of the thesis. One part is devoted to optimization in order to make methods more efficient and produce shorter paths. In the end of the thesis are described experiments and their results with different number of places from 101 up to 3891.
|
|
|
Word2vec Models with Added Context Information
Šůstek, Martin ; Rozman, Jaroslav (referee) ; Zbořil, František (advisor)
This thesis is concerned with the explanation of the word2vec models. Even though word2vec was introduced recently (2013), many researchers have already tried to extend, understand or at least use the model because it provides surprisingly rich semantic information. This information is encoded in N-dim vector representation and can be recall by performing some operations over the algebra. As an addition, I suggest a model modifications in order to obtain different word representation. To achieve that, I use public picture datasets. This thesis also includes parts dedicated to word2vec extension based on convolution neural network.
|
|
|
Overview of Actual Approaches to Optimization
Hudecová, Patrícia ; Šůstek, Martin (referee) ; Zbořil, František (advisor)
This work aimed to study some of the optimization algorithms inspired by nature and to test their success in finding the extreme of a function on various functions. Four algorithms were selected, namely the bat algorithm, the firefly algorithm, the flower pollination algorithm, and the black hole algorithm. The Griewank function, the Rastringin function, and the Rosenbrock function were chosen as test functions for finding the extreme of the function. The work contains a description of individual algorithms, a description of test functions and a description of the experiments, and an evaluation of the success of the algorithms.
|
|
|
Neural Network Based Image Modifications
Maslowski, Petr ; Zbořil, František (referee) ; Šůstek, Martin (advisor)
This thesis deals with image colorization and image super-resolution using neural networks. It briefly explains neural networks principles and summarizes current approaches in this domain. It also describes the design, implementation and training of various neural network architectures. The best implemented architecture can colorize images, in particular, works well with outdoor areas. The architecture for image super-resolution with residual blocks that was trained with a perceptual loss function performs a double increase in image resolution (4x more pixels in total). Part of this thesis is also an implementation of a web application that uses trained models for image modification.
|
|
|
Film Suggestions Based on CSFD User Profiles
Janko, Pavel ; Šůstek, Martin (referee) ; Uhlíř, Václav (advisor)
This thesis covers the topic of utilizing neural nets for recommending movies. The principle of using neural nets with machine learning and both the general and the advanced techniques of creating a recommender system are also covered in the thesis. The core of the thesis is the design, implementation and finally the evaluation of a system for movie recommendations based upon the data mined from the user profiles from the ČSFD (Czech-Slovak film database). In order to accomplish this goal the system utilizies an explicit factorization model based on collaborative filtering between items to predict an accurate rating that the user would presumably give to a movie after watching it. This thesis also describes the relation between dataset size and prediction accuracy and demonstrates this accuracy by analyzing user feedback.
|
|
|
Overview of Actual Approaches to Classifications
Brezánský, Tomáš ; Šůstek, Martin (referee) ; Zbořil, František (advisor)
This bachelor thesis deals with an overview of current approaches to classifications. It describes various approaches to classifications and their algorithms, focuses on neural networks, Bayesian classifiers and decision trees. The main task of this work is to perform experiments with three classification algorithms, namely, the ID3 algorithm, the RCE neural network and the naive Bayesian classifier. The work contains experiments with given algorithms and evaluates the obtained results.
|
guest ::
RSS feed.