|
|
Deep Learning for Image Stitching
Držíková, Diana Maxima ; Vaško, Marek (oponent) ; Španěl, Michal (vedoucí práce)
Stitching digital images is not something unfamiliar to the average technology user. The most common example of stitching can be found in panoramic images, where the algorithm stitches them to achieve a seamless, high-quality picture. Various steps need to be executed to stitch the images. Feature detection, description, and matching play the most important role in achieving the goal. This thesis will dwell deeper into the stitching problematic and will discuss the possible solutions. The traditional approaches to stitching will be explained in order to understand the basic idea behind it. Later on, the neural networks will be used to enhance the feature processing. The SuperPoint and SuperGlue neural networks will be discussed and used for their experiments. The main product of this work is a matching algorithm which uses the SuperPoint and SuperGlue models to stitch the images from grids. Other experiments which helped the process of understanding this problem, will be explained and evaluated.
|
|
|
Iris Image Quality Assessment
Vaško, Marek ; Herout, Adam (oponent) ; Hradiš, Michal (vedoucí práce)
Iris image recognition is one of the most accurate ways of biometric identification. Various verification errors can be caused if the biometric system receives poor input. By assessing the image quality it is possible to eliminate inputs causing such errors. There is a relatively insignificant development in the field of iris quality assessment and many methods that could potentially be used have not been tested in this area yet. This work focuses on different quality assessment methods used in face recognition. These quality assessment methods are then applied to the area of iris identification. The solution uses verification systems based on various iResNet and MobileNetV3 architectures. Selected quality assessment methods are applied to individual systems. Different quality assessment methods train either the system directly or use its outputs to obtain information about quality. The resulting system achieves a reduction of false non-match rate by up to 56% with the absolute value of 0.5% for iResNet50 and up to 22 \% with the absolute value of 6.4% for MobileNetV3 when using the best quality assessment method. The results are given for the data set University of Notre Dame Iris CrossSensor 2013 with an input reject rate of 10% and a false match rate of 0.1%.
|
|
|
Autonomous Rover Navigation on Planetary Surface
Vaško, Marek ; Prustoměrský, Milan (oponent) ; Chudý, Peter (vedoucí práce)
Exploration of the depths of the space has led to the development of technology in various fields. One of these areas is the exploration of the surface of extraterrestrial planets. An unmanned ground vehicle is an effective way of exploration. This thesis deals with one of the most important systems of unmanned vehicles, which is autonomous navigation. The vehicle must be able to navigate in the environment and map potential obstacles. The thesis will examine the navigation principles that have been used by existing vehicles. It will then research the use of the algorithm based on the principle of simultaneous localization and mapping and its implementation in MATLAB. This algorithm will be integrated into the simulator, which will allow later integration into the real environment using the Robot Operating System. A vehicle platform with simulated sensors and a six-wheel chassis, which will be used for an integrated algorithm, will be designed in the simulator. Finally, the quality of the proposed algorithm is evaluated and a discussion about future improvements is initiated.
|
|
|
Identifikace osob pomocí obrazu duhovky
Žákovic, Marek ; Hradiš, Michal (oponent) ; Vaško, Marek (vedoucí práce)
The goal of this bachelor’s thesis was to create a system for person identification using iris images. The thesis describes existing methods and procedures for iris recognition. The proposed method utilizes a convolutional neural network trained to extract features, which are then used to compare whether the image belongs to the same person or not. The experiments involve training and evaluating the neural network. For the purposes of this thesis, freely available datasets were used, which were modified for specific use.
|
|
|
Automatizované hodnocení kvality snímků sítnice pomocí strojového učení
Mikheda, Vladislav ; Vaško, Marek (oponent) ; Kavetskyi, Andrii (vedoucí práce)
Tato práce se zaměřuje na řešení problému hodnocení kvality snímků sítnice. Při diagnostice onemocnění lékaři se zaměřují na kvalitu jednotlivých anatomických struktur sítnice, podle kterých probíhá diagnostika. Cílem této práce je navrhnout a implementovat program pro automatizované hodnocení kvality snímků sítnice na základě anatomických struktur pomocí neuronových sítí. K řešení výše uvedeného problému bylo vyvinuto a implementováno celkem šest neuronových sítí. Tři z nich měly za úkol segmentovat jednotlivé anatomické struktury sítnice a tři další měly vyhodnocovat snímky na základě kvality segmentované struktury. Bylo provedeno jak testování každé neuronové sítě zvlášť, tak testování celého programu. Model umožňuje hodnocení kvality snímků sítnice na základě anatomických struktur.
|
|
|
Aplikace pro rozpoznávání nemocí rostlin
Kozub, Tadeáš ; Vaško, Marek (oponent) ; Bažout, David (vedoucí práce)
Práce se zabývá návrhem a tvorbou mobilní aplikace, která slouží k rozpoznávání nemocí rostlin. Věnuje se návrhu aplikace, jejího uživatelského rozhraní a jeho testování, tvorbě REST API a také trénování modelů pomocí metod strojového učení. Výsledkem práce je aplikace vytvořená v React Native, vytrénovaný model sloužící k rozpoznávání nemocí z fotografií listů rajčat a implementace backendu aplikace hostovaná na vzdáleném serveru. Model pro rozpoznávání nemocí rajčat dosáhl úspěšnosti 98,76 %.
|
|
|
Digitization of Handwritten Chess Game Sheets
Šiška, Krištof ; Vaško, Marek (oponent) ; Španěl, Michal (vedoucí práce)
Chess is one of the most popular board games in the world. An enormous amount of chess games are played daily and its popularity is still on the rise. When playing live chess games, transcripts of the chess matches are created as chess records, also known as chess score sheets. Transcribing these score sheets into digital format is a tedious and time-consuming task. The time spent on transcription increases exponentially if the handwriting is illegible or if the game contains a large number of moves. This work focuses on the problem of transcribing chess score sheets into digital format and reducing the amount of time spent by humans on this necessary but often tedious task in many areas.
|
|
|
Detekce čárových kódů v obraze
Vašíček, Vojtěch ; Vaško, Marek (oponent) ; Nguyen, Son Hai (vedoucí práce)
Cílem této práce je zhodnocení současných metod detekce čárových kódů v obraze, otestování několika vybraných metod a následné navržení a implementace vylepšení některé z těchto metod. V této práci byla otestována Scharrova metoda detekce, detektor knihovny OpenCV a neuronová síť YOLOv5. Samotné vylepšení bylo provedeno kombinací neuronové sítě YOLOv5 a detektoru čárových kódů knihovny OpenCV. Podařilo se dosáhnout zlepšení average precision při prahu 0.5 o 3.42% oproti metodě YOLOv5 a 38.38% oproti detektoru čárových kódů knihovny OpenCV.
|
|
|
Deep Learning for Image Stitching
Držíková, Diana Maxima ; Vaško, Marek (oponent) ; Španěl, Michal (vedoucí práce)
Stitching digital images is not something unfamiliar to the average technology user. The most common example of stitching can be found in panoramic images, where the algorithm stitches them to achieve a seamless, high-quality picture. Various steps need to be executed to stitch the images. Feature detection, description, and matching play the most important role in achieving the goal. This thesis will dwell deeper into the stitching problematic and will discuss the possible solutions. The traditional approaches to stitching will be explained in order to understand the basic idea behind it. Later on, the neural networks will be used to enhance the feature processing. The SuperPoint and SuperGlue neural networks will be discussed and used for their experiments. The main product of this work is a matching algorithm which uses the SuperPoint and SuperGlue models to stitch the images from grids. Other experiments which helped the process of understanding this problem, will be explained and evaluated.
|
|
|
Iris Image Quality Assessment
Vaško, Marek ; Herout, Adam (oponent) ; Hradiš, Michal (vedoucí práce)
Iris image recognition is one of the most accurate ways of biometric identification. Various verification errors can be caused if the biometric system receives poor input. By assessing the image quality it is possible to eliminate inputs causing such errors. There is a relatively insignificant development in the field of iris quality assessment and many methods that could potentially be used have not been tested in this area yet. This work focuses on different quality assessment methods used in face recognition. These quality assessment methods are then applied to the area of iris identification. The solution uses verification systems based on various iResNet and MobileNetV3 architectures. Selected quality assessment methods are applied to individual systems. Different quality assessment methods train either the system directly or use its outputs to obtain information about quality. The resulting system achieves a reduction of false non-match rate by up to 56% with the absolute value of 0.5% for iResNet50 and up to 22 \% with the absolute value of 6.4% for MobileNetV3 when using the best quality assessment method. The results are given for the data set University of Notre Dame Iris CrossSensor 2013 with an input reject rate of 10% and a false match rate of 0.1%.
|
host ::
RSS.