Name: László Marák, PhD. Faculty: Faculty of Economics and Informatics Department: Department of Informatics Position: Assistant Professor Office: E-mail: Phone +421 35 32 60 *** CV Research Publications Other informations University studies Université Marne-la-Vallée Image Segmentation. 2008 - 2007 PhD. study Université Paris-Est Image Segmentation 2008 - 2011 Employment J. Selye University Adjoint Teacher 2018 - Research area With the adoption of machine learning algorithms for image processing tasks and the ever growing need for embedded device applications, the developers use several methods to optimize the computational efficiency of their applications. Optimization of algorithms can be challenging and developers must apply non-trivial strategies to exploit the computational resources of computer architectures more efficiently. The MLP is a basic algorithm for machine learning and artificial intelligence, and is an excellent example of the difficulties surrounding GPGPU programming and optimization. As an independent observation we discuss the memory management of GPU-s and methods to simplify the memory allocation process. The optimized machine learning algorithms can have industrial applications such as in manufacturing low cost high quantity products. As there are errors during the production, it is important to be able to detect invalid products to assure that the produced products are of consistent quality. Manual quality assurance using human operators is an error prone and a costly solution. Therefore we develop an image classification method, which is using a low-cost Commercial Camera System and relies on Haar-like features combined with Maximum Relevance, Minimum Redundancy feature selection to detect the invalid products at the end of the production process. Further improving the visual acquisition pipeline, with the advent of augmented reality (AR), mixed reality (MR) and virtual reality (VR) applications, the camera angle tracking or more formally, the perspective N-point (PNP) problem is also becoming increasingly important. The accurate camera angle tracking refers to a problem of determining the position of the camera with respect to the environment in the moment of the exposition. There are some key properties of the camera which have to be known in order to provide accurate positioning. These are the intrinsic camera parameters, the distortion of the camera lens and the accurate 3D model of the environment. In our work we have developed a new camera lens calibration approach which provides a more precise and reproducible intrinsic and distortion lens calibration. We would like to specifically present two improvements: Firstly, we presented a new calibration pattern which is more immune to the calibration rig positioning problem. This problem results in uneven distribution of the calibration points on the camera lens. Secondly, we have presented an improved calibration algorithm which provides a more precise intrinsic parameter estimation and is less dependent on the actual positioning of the calibration rig, than the commonly used algorithms. As a side interest we have been exploring voting, which is an important procedure in democratic processes. The goal of our work was to propose an original electronic voting protocol. We explained the theoretical background and technical concepts of our voting protocol. We also provided formal demonstrations for our proposed algorithm. We aimed to show that we can design voting protocols that are comparable to the traditional methods of democratic elections. p { line-height: 115%; margin-bottom: 0.1in; background: transparent } Download https://www.linkedin.com/e/fpf/344458956
