|
|
Deep Learning Model Uncertainty in Medical Image Analysis
Drevický, Dušan ; Kolář, Martin (referee) ; Kodym, Oldřich (advisor)
Táto práca sa zaoberá určením neistoty v predikciách modelov hlbokého učenia. Aj keď sa týmto modelom darí dosahovať vynikajúce výsledky v mnohých oblastiach počítačového videnia, ich výstupy sú väčšinou deterministické a neposkytujú mnoho informácií o tom, ako si je model istý svojou predpoveďou. To je obzvlášť dôležité pri analýze lekárskych obrazových dát, kde môžu mať omyly vysokú cenu a schopnosť detekovať neisté predikcie by umožnila dohliadajúcemu lekárovi spracovať relevantné prípady manuálne. V tejto práci aplikujem niekoľko rôznych metrík vyvinutých v nedávnom výskume pre určenie neistoty na modely hlbokého učenia natrénované pre lokalizáciu cefalometrických landmarkov. Následne ich vyhodnotím a porovnávam v sade experimentov, ktorých úlohou je určiť, nakoľko jednotlivé metriky poskytujú užitočnú informáciu o tom, ako si je model istý svojou predpoveďou.
|
|
|
Active Learning for OCR
Kohút, Jan ; Kolář, Martin (referee) ; Hradiš, Michal (advisor)
The aim of this Master's thesis is to design methods of active learning and to experiment with datasets of historical documents. A large and diverse dataset IMPACT of more than one million lines is used for experiments. I am using neural networks to check the readability of lines and correctness of their annotations. Firstly, I compare architectures of convolutional and recurrent neural networks with bidirectional LSTM layer. Next, I study different ways of learning neural networks using methods of active learning. Mainly I use active learning to adapt neural networks to documents that the neural networks do not have in the original training dataset. Active learning is thus used for picking appropriate adaptation data. Convolutional neural networks achieve 98.6\% accuracy, recurrent neural networks achieve 99.5\% accuracy. Active learning decreases error by 26\% compared to random pick of adaptations data.
|
|
|
Parametric 2D/3D Models
Šober, Patrik ; Kolář, Martin (referee) ; Zemčík, Pavel (advisor)
The aim of this thesis was to create an application for the creation and manipulation of 2D and 3d models trough parametric construction. The work gradually describe the theoretical basis around the creation and representation of 3D models in computer graphics, the concepts of parametric construction and the introduction of similar applications, including their differences. Next is described method of creating an application based on the create design from the accuired knowledge. Finally, the application was tested, focusing on the testing user interface and the load on the rendering system.
|
|
|
Thermophoretic and Hydrodynamic Interactions in Artificial Active Matter
Kolář, Martin ; Holubec, Viktor (advisor) ; Tůma, Karel (referee)
The study of artificial active matter has significantly intensified in recent years due to its potential practical applications and promises to understand living organisms' in- teractions and collective behaviour. The research is done on all levels - experimental, theoretical, and numerical. This thesis presents a numerical simulation approach based on solving partial differential equations from the continuum mechanics framework. The results obtained can be used to predict the dynamics and thermodynamics of the system or serve as input data for stochastic simulations. Using this approach, it is also possible to simulate actual laboratory experiments numerically. 1
|
|
|
Acceleration of Particles Tracking on CBM Experiment
Roth, Michael ; Kolář, Martin (referee) ; Musil, Petr (advisor)
The focus of this work is to research various methods of particle track reconstruction in the CBM experiment, and the problem of hardware acceleration of these methods. The advantages and disadvantages of the extended methods were discussed and a reconstruction method based on cellular automata and Extended Kalman filters was selected for further study. In particular, the thesis details the development of a simulation model suitable for generating test data to facilitate the implementation of the selected tracking algorithm, which was subsequently sped up using distributed computing methods. Two different particle simulation models and a reconstruction algorithm were implemented, with the reconstruction algorithm offering up to 800 per cent speed up factor in respect to the sequential algorithm and up to four orders of magnitude lower memory complexity.
|
|
|
Acceleration of Particles Tracking on CBM Experiment
Roth, Michael ; Kolář, Martin (referee) ; Musil, Petr (advisor)
The focus of this work is to research various methods of particle track reconstruction in the CBM experiment, and the problem of hardware acceleration of these methods. The advantages and disadvantages of the extended methods were discussed and a reconstruction method based on cellular automata and Kalman filters was selected for further study. In particular, the thesis details the development of a simulation model suitable for generating test data to facilitate future implementation of the selected tracking algorithm. Two different particle simulators have been developed and will be used in the following work to calculate the prediction step of the extended Kalman filter and to test the quality of the implemented reconstruction method.
|
|
|
Fusion of Radar and Visual Data for Remote Sensing
Strych, Tomáš ; Beran, Vítězslav (referee) ; Kolář, Martin (advisor)
Cieľom práce je vygenerovať satelitný optický snímok v prípade jeho nedostupnosti. Takýto snímok je vygenerovaný z aktuálneho radarového snímku a za pomoci radarových a optických snímkov z minulosti. Zameranie práce cieli na poľnohospodárstvo, kde sa na analýzu dát používajú rôzne vegetačné indexy. Pre zjednodušenie problematiky je práca zameraná len na optický snímok zobrazujúci NDVI. Boli vytvorené 4 dátové sady, pre prvé tri ročné obdobia a~štvrtý, ktorý ich spája. Pre riešenie problému preloženia obrázku z jedného na druhý bol použitý model Pix2Pix-cGAN. Výsledky práce zobrazujú rozdiely pri použití dátových sád, rozličného množstva a~typu použitých snímkov, tak ako aj intervalu medzi snímkami. Daným výskumom bolo zistené, že sieť je schopná vytvárať reálne uveriteľné obrázky s validnými numerickými hodnotami, avšak má problém správne využiť informáciu o radarovej zmene, ktorá je dôležitá pre ohodnotenie vývoja rastlín práve v prípade nedostupnosti optického snímku. Táto práca a~jej výsledky sú jedinečné vďaka naprieč Európou geograficky rozmanitej dátovej sade a vďaka zameraniu na agrikultúru, a to bez ohľadu na typ plodín.
|
|
|
Computer Vision with Active Learning
Kolář, Martin ; Machová, Kristína (referee) ; Arridge, Simon (referee) ; Zemčík, Pavel (advisor)
Metody strojového vidění se zdokonalují zlepšením modelů, laděním trénovaných parametrů nebo anotací reprezentativních dat. Tato práce řadou experimentů potvrzuje hypotézu že aktivní učení zvyšuje přesnost těchto modelů. Rozšířením přistupu pseudolabelů o aktivní učení přispívá tato práce přístupem "one-shot-learning" k učení nových kategorií obrazů s použitím algoritmických doporučení, dále online grafickým uživatelským rozhraním pro optimalizaci dilema Exploration/Exploitation pro online tagování, a dvoukrokovým offline binárním přístupem aktivního učení pro zlepšení kvality dat používaných pro snímání fontů. Tím, že demonstruje přínos aktivního učení v těchto přístupech, přispívá tato práce k hypotéze i konkrétním aplikacím strojového vidění.
|
|
|
Prevention and importance of ventilator associated pneumonia in the intensive care
Sekerka, Pavel ; Kolář, Martin (advisor)
Synopsis Ventilation-associated pneumonia (VAP) presents a severe complication in the care for the mechanically-ventilated patient and its importance rises with the increasing number of resistant microorganisms, which causes VAP. VAP increases morbidity and mortality of the patients, prolongs their hospitalization and leads to the increase of direct and indirect treatment expenses. Options of treatment are sometimes difficult, costly and inefficient, thereby the need to prevent the occurance of VAP by employing a set of preventive measures is needed. Employing and enforcing those measures can decrease the incidence of VAP and thus to improve prognosis of the critically-ill patients in the intesive care. The goal of the experimental part of this thesis was to audit the employment of these preventive measures in practice. A survey using questionnaires and observing the situation on KAR FNKV have shown, that some measures are performed correctly, while some present a space for improvement, some are performed incorrectly or not at all.
|
|
|
How do objects arise?
Novotný, Michal ; Novák, Aleš (advisor) ; Kolář, Martin (referee)
Proposed thesis is an attempt to summarise the term object in the work of American philosopher Graham Harman, he himself named Object oriented ontology. The accent will be given also to other terms such as realism, essence, quadruple dimension of object, withdrawal of object, intentionality, conception of time and space, emanation and others. It attempts to create a certain genealogy of those terms in the work of other authors (Kant, Schopenhauer, Nietzche, Merleau-Ponty) and consequent comparison with the use of those terms in Harman's work. The aim is to create a certain critical overview of Harman's conception of o object and in general speculative realism theory and point out certain other approached to the classical subject-object dimension.
|
guest ::
RSS feed.