|
Capturing of Detailed and Very Large Photograph and Localization Within
Dubovec, Pavol ; Vaško, Marek (referee) ; Herout, Adam (advisor)
The goal of this work was to create a large image and a new technique to localize the photo in the larger image to increase the speed and accuracy of conventional methods. The proposed technique uses CNN architecture to extract embeddings from the queried image which will be used to search the database of embeddings from the large photo. Two models have been trained on a large dataset: based on classification (CE) and distance (triplet) Conventional methods were used to determine the location of the images and to generate the large image. A database of embeddings was created by partitioning the large image using the trained model. The database is searched for the K-nearest embeddings of the cutouts of the query image. These embeddings are generated by dividing the query photo into the same size parts as the CNN inputs. The optimal homography model is determined by random selection based on the positions of the four query image cutouts and their corresponding positions in the big picture. The homography model with the lowest harmonic mean of the embedding distance is selected as the final position. The homography is optimized using template matching where possible. The method shows sufficient accuracy and high speed on test datasets. The best model achieved a top-1 accuracy of 97.71% and a top-3 accuracy of 99.67%. Future research will investigate the performance of the method under increasing surface heterogeneity, the possibility of automating video retrieval to obtain a large dataset with photos, and its effectiveness in locating photos when conventional methods fail.
|
|
Chatbot Capable of Information Search
Ďurista, Michal ; Beneš, Karel (referee) ; Černocký, Jan (advisor)
Pojem ''chatbot'' je v dnešnej dobe umelej inteligencie veľmi populárny výraz. Chatbotov vidno stále viac a viac v biznis riešeniach dnešných firiem. Hlavným cieľom práce je vytvoriť algoritmus, ktorý je schopný vyťahovať informácie a implementovať ho do chatbota. Tieto informácie možno nájsť na webových stránkach reálneho zákazníka. Práca rovnako poskytuje prehľad súčasnej situácie chatbotov ako aj Microsoft technológií pre ich vývoj. Technické detaily na ktorých tieto technológie pracujú, predovšetkým spracovanie prirodzeného jazyka, sú taktiež zahrnuté. Práca popisuje implementáciu algoritmu ako aj chatbota samotného spolu s procesom testovania v skutočnom priemyselnom prostredí.
|
|
The Comparison of the Algorithms for the Solution of Travel Sales Problem
Kopřiva, Jan ; Všetička, Martin (referee) ; Dostál, Petr (advisor)
The Master Thesis deals with logistic module innovation of information system ERP. The principle of innovation is based on implementation of heuristic algorithms which solve Travel Salesman Problems (TSP). The software MATLAB is used for analysis and tests of these algorithms. The goal of Master Thesis is the comparison of selections algorithm, which are suitable for economic purposes (accuracy of solution, speed of calculation and memory demands).
|
|
Domain Specific Data Crawling for Language Model Adaptation
Gregušová, Sabína ; Švec, Ján (referee) ; Karafiát, Martin (advisor)
The goal of this thesis is to implement a system for automatic language model adaptation for Phonexia ASR system. System expects input in the form of source that, which is analysed and appropriate terms for web search are chosen. Every web search results in a set of documents that undergo cleaning and filtering procedures. The resulting web corpora is mixed with Phonexia model and evaluated. In order to estimate the most optimal parameters, I conducted 3 sets of experiments for Hindi, Czech and Mandarin. The results of the experiments were very favourable and the implemented system managed to decrease perplexity and Word Error Rate in most cases.
|
|
Graphic Animation of Problem Solving Methods
Macek, Jiří ; Jurka, Pavel (referee) ; Zbořil, František (advisor)
There are many kinds of implementation artificial intelligence for automatic solving problems by computer technology. The main topics of this bachelor's thesis are some typical methods, describing of their features, comparing them among and shows some useful techniques of algoritmization and implementation too. Main purpose of this thesis is creating application, which clearly demonstrates at chosen problems methods of their solving.
|
|
Search of Metric State Space with Obstacles
Lukáč, Jakub ; Rozman, Jaroslav (referee) ; Šůstek, Martin (advisor)
This thesis is focused on search of metric state space with obstacles. The thesis selects four methods based on state space search and presents two new algorithms, which will try to consider an obstacles in metric space. The selected methods and the new designed algorithms are implemented as an application in programming language Java, application is also part of the thesis. The thesis presents experiments on metric spaces with various kind of obstacles for comparison of individual methods.
|
|
Searching Acoustic Patterns in Speech Data without Recognition
Skácel, Miroslav ; Fapšo, Michal (referee) ; Černocký, Jan (advisor)
This work investigates into methods for words, word phrases and longer segments detection in large speech data sets in an unsupervised way. At first, basics for the given topic and principles of modern methods for searching of repeating objects are introduced. The representation and segmentation of the input data are described. Techniques for object detection in speech are presented. The description of found motifs modelling follows. The next step defi nes data sets for experiments in which spoken term detection by an example is performed. The system requirements are described. In the conclusion, the work is summarised and suggestions for further development are discussed.
|
|
A Comparison Particle Filter for Searching a Radiation Source in Real and Simulated World
Cihlar, Milos ; Lazna, Tomas ; Zalud, Ludek
In this paper, we are focusing on comparing solutionsfor localizing an unknown radiation source in both aGazebo simulator and the real world. A proper simulation ofthe environment, sensors, and radiation source can significantlyreduce the development time of robotic algorithms. We proposeda simple sampling importance resampling (SIR) particle filter.To verify its effectiveness and similarities, we first tested thealgorithm’s performance in the real world and then in the Gazebosimulator. In experiment, we used a 2-inch NaI(Tl) radiationdetector and radiation source Cesium 137 with an activity of 330Mbq. We compared the algorithm process using the evolution ofinformation entropy, variance, and Kullback-Leibler divergence.The proposed metrics demonstrated the similarity between thesimulator and the real world, providing valuable insights toimprove and facilitate further development of radiation searchand mapping algorithms.
|
|
Domain Specific Data Crawling for Language Model Adaptation
Gregušová, Sabína ; Švec, Ján (referee) ; Karafiát, Martin (advisor)
The goal of this thesis is to implement a system for automatic language model adaptation for Phonexia ASR system. System expects input in the form of source that, which is analysed and appropriate terms for web search are chosen. Every web search results in a set of documents that undergo cleaning and filtering procedures. The resulting web corpora is mixed with Phonexia model and evaluated. In order to estimate the most optimal parameters, I conducted 3 sets of experiments for Hindi, Czech and Mandarin. The results of the experiments were very favourable and the implemented system managed to decrease perplexity and Word Error Rate in most cases.
|
|
GPU Accelerated Adversarial Search
Brehovský, Martin ; Bošanský, Branislav (advisor) ; Bída, Michal (referee)
General purpose graphical processing units were proven to be useful for accelerating computationally intensive algorithms. Their capability to perform massive parallel computing significantly improve performance of many algorithms. This thesis focuses on using graphical processors (GPUs) to accelerate algorithms based on adversarial search. We investigate whether or not the adversarial algorithms are suitable for single instruction multiple data (SIMD) type of parallelism, which GPU provides. Therefore, parallel versions of selected algorithms accelerated by GPU were implemented and compared with the algorithms running on CPU. Obtained results show significant speed improvement and proof the applicability of GPU technology in the domain of adversarial search algorithms.
|