|
|
Crowd Counting in Video
Kuřátko, Jiří ; Sochor, Jakub (referee) ; Herout, Adam (advisor)
This master's thesis prepared the programme which is able to follow the trajectories of the movement of people and based on this to create various statistics. In practice it is an effective marketing tool which can be used for instance for customer flow analyses, optimal evaluation of opening hours, visitor traffic analyses and for a lot of other benefits. Histograms of oriented gradients, SVM classificator and optical flow monitoring were used to solve this problem. The method of multiple hypothesis tracking was selected for the association data. The system's quality was evaluated from the video footage of the street with the large concentration of pedestrians and from the school's camera system, where the movement in the corridor was monitored and the number of people counted.
|
|
|
Mobile App for Scanning and Recognition of Cashier Bills
Bambuch, Vladislav ; Sochor, Jakub (referee) ; Herout, Adam (advisor)
The aim of this thesis is to develop a mobile application which allows photographing of cashier bills and offers a function of finance management via an intuitive graphical user interface. The app uses OCR technology to digitize the prices of receipts and follows the new Android Material Design trends. The price marking of the particular items is carried out on the taken receipt. When the user indicates particular prices by clicking on them they are summed up into the final amount. Users are able to save such final amounts together with other related information and record statistics about their spending. That results in a fast and very easy way of adding the purchase records which are available both in form of organized summaries and graphs.
|
|
|
Apparent Personality Analysis from Video
Čigáš, Patrik ; Sochor, Jakub (referee) ; Hradiš, Michal (advisor)
This bachelor thesis deals with experiments with systems for apparent personality analysis from video, and compares accuracy of these systems. Systems from the experiments are created by linear regression and convolutional neural networks. Experiments compare accuracy of linear regressors processing visual and audial modality of video. On spectograms made from audial modality of video, thesis evaluates results of convolutional neural networks with varying number of convolutional and fully connected layers nad subsequently compares accuracy of regression solution and classification solution of the problem. For visual modality of video the thesis compares information values of gaze movement and face landmarks movement. System processing face landmarks movement reaches the best results in the experiments.
|
|
|
Image Compression with Neural Networks
Teuer, Lukáš ; Sochor, Jakub (referee) ; Hradiš, Michal (advisor)
This document describes image compression using different types of neural networks. Features of neural networks like convolutional and recurrent networks are also discussed here. The document contains detailed description of various neural network architectures and their inner workings. In addition, experiments are carried out on various neural network structures and parameters in order to find the most appropriate properties for image compression. Also, there are proposed new concepts for image compression using neural networks that are also immediately tested. Finally, a network of the best concepts and parts discovered during experimentation is designed.
|
|
|
Reconstruction of 3D Information about Vehicles Passing in front of a Surveillance Camera
Dobeš, Petr ; Sochor, Jakub (referee) ; Herout, Adam (advisor)
This master's thesis focuses on 3D reconstruction of vehicles passing in front of a traffic surveillance camera. Calibration process of surveillance camera is first introduced and the relation of automatic calibration with 3D information about observed traffic is described. Furthermore, Structure from Motion, SLAM, and optical flow algorithms are presented. A set of experiments with feature matching and the Structure from Motion algorithm is carried out to examine results on images of passing vehicles. Afterwards, the Structure from Motion pipeline is modified. Instead of using SIFT features, DeepMatching algorithm is utilized to obtain quasi-dense point correspondences for the subsequent reconstruction phase. Afterwards, reconstructed models are refined by applying additional constraints specific to the vehicle reconstruction task. The resultant models are then evaluated. Lastly, observations and acquired information about the process of vehicle reconstruction are utilized to form proposals for prospective design of an entirely custom pipeline that would be specialized for 3D reconstruction of passing vehicles.
|
|
|
Mobile Application for Studis BUT
Smyčka, Jan ; Sochor, Jakub (referee) ; Špaňhel, Jakub (advisor)
Bachelor thesis deals with the development of mobile application in Android for VUT information system. The theoretical part of the thesis describes Android operating system and its basic elements used in the application development process, the text also describes the architectural design of mobile applications. The main part of the thesis deals with the implementation of the application and its testing on the users. In the final part of the thesis, the application development and user testing are assessed, and further possible development of the application for the future is proposed.
|
|
|
On-Board License Plate Detection and Recognition
Tomovič, Martin ; Sochor, Jakub (referee) ; Špaňhel, Jakub (advisor)
This Bachelor's thesis aims to create an aplication for detection and recognition of license plates suitable for real-time processing. The work contains analysis of available methods. Part of the work is focused on present form of licence plates in Czech Republic. As a result of work, new data set was created and computer application was implemented. The application uses existing libraries designed for computer vision and machine learning with main purpose to detect and recognize licence plates from video. Detection is realized with help of cascade classifier, and recognition by Perceptron neural network. Final chapter subsequently contains evaluation of success rate of implemented solution.
|
|
|
Automatic Traffic Video Surveillance: Fine-Grained Recognition of Vehicles and Automatic Speed Measurement
Sochor, Jakub ; Elder, James (referee) ; Svoboda,, Tomáš (referee) ; Herout, Adam (advisor)
V rámci této dizertační práce se zaměřuji na Inteligentní dopravní systémy a Počítačové vidění - především automatické měření rychlosti a rozpoznání automobilů podle typů. Rozpoznání automobilů podle typů je úkol, ve kterém system má predikovat přesný typ (např. Škoda Octavia combi mk2) pro daný obrázek automobilu. Publikoval jsem dva články, které popisují navržený přístup k tomuto problému a tvoří jádro této dizertace. Prezentovaná metoda je založena na 3D obalových kvádrech postavených okolo automobilů, které jsou následně využity pro rozbalení obrázku automobilu do roviny a tudíž normalizaci vstupu neuronové sítě, která dělá následné rozpoznání. Přístup byl dále rozpracován v druhé publikaci, kde je navržena metoda pro určení tohoto 3D obalového kvádru z jediného obrázku - tudíž není nutné mít zkalibrovanou kameru. Experimentální výsledky ukazují, že navržená metoda zlepšuje úspěšnost rozpoznání o 12 procentních bodů - chyba rozpoznání je redukována o 50 procent.Při měření rychlosti má systém za úkol odhadnout rychlost projíždějících aut z videa. Cílem je také, ať měření probíhá plně automaticky bez jakékoli manuální kalibrace. Jelikož neexistoval žádný dataset, který by obsahoval velké množství průjezdů s přesně změřenou rychlostí, tak jsme nejprve takovýto dataset pořídili. Dále jsem navrhnul metodu pro plně automatickou kalibraci dopravní dohledové kamery což umožňuje měřit rychlost automobilů pozorovaných touto kamerou. Metoda je založena na odhadu kalibrace pomocí detekovaných úběžníků scény a následného zarovnání 3D modelů několika běžných typů automobilů. Experimentální výsledky ukazují, že navržená metoda dosahuje průměrné chyby měření rychlosti 1,10 km/h.
|
|
|
Fine-Grained Vehicle Recognition from Traffic Surveillance Camera
Mencner, Pavel ; Špaňhel, Jakub (referee) ; Sochor, Jakub (advisor)
The aim of this thesis is image based detection of vehicles from traffic surveillance camera and fine-grained vehicle type recognition (manufacturer and model). In the thesis the Unpack normalization method is implemented which transforms the vehicle image into its apparent flat representation in order to increase the classifier's success rate. The Unpack method make use of 3D bounding box of the vehicle. This bounding box is constructed during test period using the information of vehicle contour and direction toward vanishing points. The thesis involve accuracy comparison between direct and Unpack classification methods. The proposed solution is based on several related parts that benefit from convolutional neural networks. These parts are: vehicle detection from image data, estimation of the directions towards vanishing points solved as classification task, vehicle contour detection using convolutional Encoder-Decoder network and fine-grained vehicle type classification. Using Unpack based classification the 2% accuracy improvement against direct classification has been achieved, resulting in 86% overall success rate. The outcome of this thesis is fine-grained vehicle classification system that works with traffic surveillance video without any viewpoint limitations.
|
|
|
License Plate Detection and Recognition from Still Image
Janíček, Kryštof ; Sochor, Jakub (referee) ; Špaňhel, Jakub (advisor)
This thesis describes the design and implementation of system for detection and recognition of license plate. This system is divided into three parts which are license plate detection, character segmentation and optical character recognition. License plate detection is done by cascade classifier that achieves hit rate of 95.5% and precision rate of 95.9%. Character segmentation is based on contour finding that achieves hit rate of 93.3% and precision rate of 96.5%. Optical character recognition is done by neural network and achieves hit rate of 98.4% for individual characters. The whole system is able to detect and recognize up to 81.5% of license plates from the test data set.
|
guest ::
RSS feed.