|
|
Semantic Segmentation in Mountainous Environment
Pelikán, Jakub ; Čadík, Martin (referee) ; Brejcha, Jan (advisor)
Semantic segmentation is one of classic computer vision problems and strong tool for machine processing and understanding of the scene. In this thesis we use semantic segmentation in mountainous environment. The main motivation of this work is to use semantic segmentation for automatic location of geographic position, where the picture was taken. In this thesis we evaluated actual methods of semantic segmentation and we chose three of them that are appropriate for adapting to mountainous environment. We split the dataset with mountainous environment into validation, train and test sets to use for training of chosen semantic segmentation methods. We trained models from chosen methods on mountainous data. We let segments from the best trained models get evaluated in electronic survey by respondents and we evaluated these segments in process of camera orientation estimation. We showed that chosen methods of semantic segmentation are possible to use in mountainous environment. Our models are trained on 11, 5 or 4 mountainous classes and the best of them achieve on 4 class mean IU 57.4%. Models are usable in practise. We show it by their deployment as a part of camera orientation estimation process.
|
|
|
Visual Localization in Natural Environments
Brejcha, Jan ; Sattler, Torsten (referee) ; Matas, Jiří (referee) ; Čadík, Martin (advisor)
V této práci se zabýváme odhadem pozice a orientace kamery z dané fotografie. Tento problém nazýváme vizuální geo-lokalizace. Konkrétně se zabýváme fotografiemi pořízenými v přírodních horských prostředích. Představujeme podrobný průzkum aktuálního stavu poznání algoritmů, datových sad a přístupů k vyhodnocování problému vizuální geo-lokalizace. Náš průzkum odhalil, že vizuální geo-lokalizace v přírodních prostředích je často řešena pomocí vyhledávání podobností nebo korespondencí mezi vstupní fotografií a terénním modelem. Problém nacházení korespondencí mezi fotografií a terénním modelem nazýváme porovnávání napříč doménami (cross-domain matching). Na základě našeho průzkumu jsme stanovili tři hlavní cíle, jejichž dosažení nám umožňuje překonat aktuální stav poznání vizuální geo-lokalizace v horských prostředích s využitím porovnávání napříč doménami: (I) potřeba nových datových sad které umožní trénovat, vyhodnocovat a porovnávat algoritmy vizuální geo-lokalizace, (II) potřeba ověřit, zda využití různých příznaků - křivek horizontu, hranových map, sémantické segmentace a satelitních snímků pomůže vylepšit algoritmy pro porovnávání napříč doménami, (III) potřeba ilustrovat využitelnost metod vizuální geo-lokalizace pomocí vývoje jejich nových aplikací. V této práci podrobně popisujeme naše výzkumné studie, které objasňují, jakým způsobem jsme postupovali ve výzkumu jednotlivých cílů. Představujeme několik nových datových sad pro účely vyhodnocování, porovnávání a trénování jednotlivých metod. S využitím těchto nových datových sad jsme vyvinuli novou metodu pro zarovnání fotografií s terénním modelem na základě sémantické segmentace kombinované s běžnými hranovými příznaky. Pomocí experimentálního vyhodnocení objasňujeme výhody našeho nového přístupu oproti aktuálnímu stavu poznání. Dále navrhujeme meta algoritmus umožňující automatickou kalibraci více kamer, který je založen na odhadu struktury z pohybu (Structure from Motion) napříč doménami. Tento nový přístup pro automatické zarovnávání fotografií s terénním modelem nám umožňuje natrénovat kompaktní deskriptor klíčových bodů pomocí hlubokého učení. V rámci našeho výzkumu ukazujeme funkčnost tohoto deskriptoru při odhadu externích parametrů kamery (pozice a orientace) pomocí porovnávání vstupní fotografie s terénním modelem. V závěru práce ukazujeme praktickou využitelnost našich metod pro automatickou kalibraci externích parametrů kamery. Navrhujeme nový přístup k prezentaci fotografií, který je vhodný jak pro prezentaci na monitoru či jiné projekční ploše, tak pro virtuální realitu. Pomocí experimentálního vyhodnocení ukazujeme, že naše nová metoda prezentace fotografií pomáhá uživatelům s orientací v neznámých komplexních přírodních scénách.
|
|
|
Dynamic Image Presentations Using Depth Maps
Hanzlíček, Jiří ; Brejcha, Jan (referee) ; Čadík, Martin (advisor)
This master's thesis focuses on the dynamic presentation of still photography using a depth map. This text presents an algorithm that describes the process of creating a spatial model which is used to render input photography so that the movement of virtual camera creates parallax effect due to depth in image. The thesis also presents an approach how to infill the missing data in the model. It is suggested that a guided texture synthesis is used for this problem by using rendering outputs of the model themselves as guides. Additional information in model allows the virtual camera to move more freely. The final result of the camera movement can be saved to simple video sequence which can be used for presenting the input photography.
|
|
|
Image Enhancement Using Depth Maps
Krbec, Jakub ; Brejcha, Jan (referee) ; Čadík, Martin (advisor)
The thesis is dealing with methods for enhance photography captured in nature usingsynthetic depth map. Depth map is obtained from digital terrain model with precision inthe order of meters. Using depth map we can eliminate undesirable effects of weather suchas fog or we can simulate depth of field based on parameters of cameras and lenses oreventually we can simulate unreachable parameters of lenses. The method for eliminatingnegative influence of atmosphere (dehaze) is implemented by combination of available depthmap and depth map estimated directly from input photo. This combination is mainly usedif available depth map is not accurate enough. The method for simulating different depth offield is implemented by convolution of circle kernel which better approximates the shape ofthe camera aperture than Gaussian kernel. Experimental results show that using syntheticdepth map we can successfully simulate small depth of field in outdoor photos and eliminateundesirable influences of atmosphere.
|
|
|
Online System for Visual Geo-Localization in Natural Environment
Pospíšil, Miroslav ; Čadík, Martin (referee) ; Brejcha, Jan (advisor)
The goal of this master thesis is creation of an online system serving as a performing application for presentation results of visual geo-localization in nature and mountain environment. The system offers the users to choose one of the pre-defined photographs or~to~upload one's own photography while choosing a file or inserting an URL address. The~system will localizate a camera of a given image based on a visual geo-localization. The~geo-localization uses the mountain horizon as a key characteristic when searching for similar horizons. The~curve line of the horizon is extracted by a fully automatic algorithm based on supervised learning and dynamic programming. Visual geo-localization running on the server which using new inversed index with cache politic. This allows further scaling of the system. The server processing detected horizon curve and respond with set of the best candidates on results. Results are visualised to the user in form of classic map, detailed sattelite view and rendering of found panorama.
|
|
|
A Set of Web Tools for Geo-Location System
Bečka, Petr ; Brejcha, Jan (referee) ; Čadík, Martin (advisor)
The main goal of the thesis is to develop a set of web tools to visualize results and exploit functionality of geo-location system LOCATE. Obtained informations could be showen in web map and digital terrain model. For interaction with user is prepared interface to upload photos. Furthermore was developed a method, witch allows to customize appearance and functionality of visualization means.
|
|
|
Web Application for Panorama Search
Pospíšil, Miroslav ; Čadík, Martin (referee) ; Brejcha, Jan (advisor)
The aim of my bachelor thesis is to create a web front-end, which provides mountain localisation of images. Images can be uploaded directly to the application or you can use images from the Flickr web service. Flickr is integrated in the application by Flickr API. Localisation of selected image is based on mountain horizon. User-aided semi-automatic horizon detection is approached by special user interface utilising the GrabCut algorithm for segmentation. Localisation is provided by server back-end which is connected and integrated with the designed application. Results for the detected horizon are visualised in a form of aligned horizon curves with position displayed using the Google Maps API. The final fully functional application was tested by moderated usability test with users. Application in this test performed well. Despite several shortcomings detected by the test the application was easy to use.
|
|
|
Augmented Reality on iOS - Displaying of Map Features in Terrain
Bezák, Adam ; Brejcha, Jan (referee) ; Herout, Adam (advisor)
This thesis aims to improve the user's orientation in terrain using his smartphone and augmented reality. By combining these two elements and cost-free OpenStreetMap database of geographic data, is possible to show nearby map features from the maps or show additional information about nearby objects. The resulting technology is used in implemented iOS application called Water Radar. This application displays nearby water streams (or other water objects) relative to the GPS position of the user. The application also uses Google Elevation API, in order to enable displayed water streams to follow the landscape.
|
|
|
Registration of Photos to 3D Model
Deák, Jaromír ; Brejcha, Jan (referee) ; Čadík, Martin (advisor)
This work refers existing solutions and options for the task registration of photos to 3D model based on the previous knowledge of the geographic position of the camera. The contribution of the work are new ways and possibilities of the solution with the usage of graph algorithms. In this area, the work interests are useful points of interest detection in input data, a construction of graphs and graph matching possibilities.
|
|
|
Web App for Capturing New Shots of Historical Photographs
Sikora, Martin ; Brejcha, Jan (referee) ; Čadík, Martin (advisor)
The aim of this diploma thesis is to design and implement a web application focused on rephotography management. Analyze existing solutions, create list of features and simple graphical user interface. It also includes a design of API structure to communicate with the mobile application. Essential application requirements include adding photos on a map and combining different photos in a photo editor with enhanced auto-alignment features.
|
guest ::
RSS feed.