Ing. Petr Pulc, Ph.D.

Ing. Petr Pulc

Department of Software Engineering

This bachelor thesis describes a process of development of a plug-in for Adobe Premiere Pro CC. Main goal of that plug-in is to allow user to communicate with NARRA web API directly from Adobe Premiere Pro interface. User will be able to browse lists of accessible footage, import it and synchronize metadata. Result of this work is a plug-in that fulfils the requirements and can be used as a basis for future development, that provides intuitive and responsive user interface.

Thesis on DSpace

Ing. Petr Pulc

Department of Software Engineering

YouTube, perhaps the most popular server to watch and share videos worldwide. In my bachelor thesis I examine the interconnection between videos on YouTube and their labels, or metadata. Thus paired video and its metadata allows search engines in project OpenNarrative better and faster offering relevant content for further work in video, such as editing. The project is written in Ruby using the already available features of YouTube API.

Thesis on DSpace

Ing. Petr Pulc

Ing. Jan Trávníček, Ph.D.

Department of Theoretical Computer Science

This bachelor thesis deals with the classification of dance styles from music records. The thesis introduces characteristics of ballroom and latin dances, which will be searched in signals of music tracks. A relatively large part of thesis is about audio and music analysis, which contains the definition of FFT, extraction of basics and spectral descriptors and the tempo estimation algorithm. The analysis is followed by a chapter about classification, which focuses on choosing suitable searching classifier and its configuration.

Thesis on DSpace

Ing. Petr Pulc

Department of Software Engineering

This bachelor's thesis is dealing with implementation of information system for live streaming. In the theoretical part it introduces common technologies and format for audiovisual work using the Internet. In practical part is information system implemented. Used solution is compared with existing monolithic systems. System is tested and use-case models are designed. Further off is devised monetizing technique for end user content.

Thesis on DSpace

Ing. Petr Pulc

Ing. Radomír Polách

Department of Software Engineering

The topic of this work is processing records of lecture presentations. The program analyses the screen capture with presentation slides recorded with third-party programs or by recording the video signal with an external recorder. This capture is converted into a series of static slides and video files with individual slides and animations from the presentation. The file with timestamps of slides change is also included. The program is written in C++ with the support of the OpenCV library, which allows basic manipulation with videos and images.

Thesis on DSpace

Ing. Petr Pulc

Ing. Jiří Novák, Ph.D.

Department of Software Engineering

This master's thesis deals with creating a web application for user friendly creation and management of photo presentations. Emphasis is given to support of panoramic images. The thesis analyses possibility of synchronization of multiple browser windows, which is used in the application for a separate view for the presenter.

Thesis on DSpace

Ing. Petr Pulc

Ing. Jaroslav Kuchař, Ph.D.

Department of Software Engineering

This master thesis deals with an implementation of a system for retrieval, indexing and display of spherical panoramas. It describes methods of retrieval, verification and analysis of photographs from Internet. In the thesis we use methods of text mining and image processing to obtain attributes suitable for searching and clustering of panoramas.

Thesis on DSpace

Ing. Petr Pulc

Department of Software Engineering

Lending Management System With Inventorying Support is a web application for endorsing of processes related to device management, especially for devices in various laboratories. This master thesis covers replacing of several original inefficient processes by modern system which would comply with up-to-date demands placed by users. This application also supports electronic revision management. Additional functions intended for future versions are being analysed and designed.

Thesis on DSpace

Ing. Petr Pulc

Ing. Ladislav Palán, Ph.D.

Department of Applied Mathematics

Levees and embankments have a significant effect on the magnitude of flood- ing. This work deals with the identification of such structures and theirs conversion to vector GIS format. The inputs for the proposed algorithm are primarily digital terrain models; for further improvements, Sentinel-2 satellite images are used. For the detection of levees and embankments, an approach based on convolutional neural networks is proposed.

Thesis on DSpace

Ing. Petr Pulc

Department of Applied Mathematics

This thesis deals with the analysis of the possible ways of refining digital elevation models and the subsequent implementation of one of the discovered methods of solution. The method uses a convolutional neural network based on VGG16 to classify land use on a Sentinel-2 image that corresponds to the digital elevation model. Subsequently, the digital elevation model is filtered within individual classes and the method is compared with conventional filtration.

Thesis on DSpace

Ing. Petr Pulc, Ph.D.

Ing. Tomáš Vondra, Ph.D.

Department of Software Engineering

This masters thesis aims to investigate deployment options for an Apache Flink application to a cloud infrastructure on AWS. We deployed the application across three fundamental parallel computation modes: Amazon Managed Service for Apache Flink, Amazon Elastic MapReduce on EC2, and Amazon Elastic Kubernetes Service. Additionally, we conducted a cost-efficiency analysis, comparing the deployment modes. The results showed that the Amazon Managed Service for Apache Flink is the most cost-efficient deployment mode for the given application. For more complex applications with higher computational requirements or faster processing speed, Amazon Elastic Kubernetes Service (EKS) proved to provide a more economical solution. We used Terraform for the deployment of the Flink application, as well as other components for additional infrastructure. This approach allows for the easy reproducibility of our findings and customizing deployments further to suit the readers needs. We also implemented tests to validate the Terraform deployment and conducted tests to verify the output consistency. This thesis contributes to providing deep insights for users intending to deploy Big Data processing applications to the cloud, including actual cost analysis and comparing the three deployment modes. All of this should simplify otherwise complex decision-making and cost-estimation processes.

Thesis on DSpace