|
|
KNOTIS Information System
Ľupták, Andrej ; Smrž, Pavel (referee) ; Dytrych, Jaroslav (advisor)
This thesis addresses the problem of managing people, groups of people, resources, and resource permissions in the Unix system managed by the Knowledge Technology Research Group Information System (KNOTIS). The main objective is to effectively combine the use of access lists and groups to manage user access to resources. Based on the properties of the resource and the number of persons with access, the module automatically evaluates which type of permissions is more advantageous for managing permissions to individual resources. The communication between KNOTIS and a server is implemented using the JSON-RPC communication protocol. The content of the communication messages was defined based on specific demands. The result of this thesis is a new group of modules for KNOTIS and the servers managed by it. The modules focus on a new way of communication between the information system and the servers and on how to process the requested changes on the server side, including a more efficient way of managing resource permissions.
|
|
|
Information Extraction from Wikipedia
Jurišica, Rudolf ; Otrusina, Lubomír (referee) ; Smrž, Pavel (advisor)
The goal of this thesis is to reduce the number of unknown referenced entities in Czech Wikipedia articles. This has been achieved by using some existing solutions, created by the KNOT research group at FIT BUT, and then by creating a set of programs. These programs are automatically run every month, when a new version of Wikipedia is released. They will automatically add new names to the knowledge base, generate their derived forms, and edit the articles themselves directly on Wikipedia.
|
|
|
Intelligent Environment for Extending Python Programming Knowledge via Self-learning
Krejčí, Jan ; Nosko, Svetozár (referee) ; Smrž, Pavel (advisor)
This thesis aims to create an intelligent environment for extending the knowledge of Python programming through self-study. A key element of this work is the implementation of feedback mechanisms. For this purpose, the capabilities and limitations of large language models have been analyzed. The developed system uses classification models to provide personalized feedback based on the analysis of student projects. The system has been deployed and tested in the Scripting Languages course at FIT BUT and received positive feedback from students. The outcome presents a comprehensive and functional system that has fulfilled its original intention and contributed to a more effective and interactive Python programming education process.
|
|
|
Large Language Models for Generating Code Focusing on Embedded Systems
Vadovič, Matej ; Nosko, Svetozár (referee) ; Smrž, Pavel (advisor)
The goal of this work was to adapt a pre-trained language model for the purpose of generating code in the field of embedded systems. The work introduces a new dataset for fine-tuning code generation models, consisting of 50,000 pairs of source code and comments focused on embedded systems programming. This dataset is composed of collected source code from the GitHub platform. Two new language models for code generation, based on transformer architecture pre-trained models, were fine-tuned on the data of the new corpus. Model MicroCoder is based on the CodeLLaMA-Instruct 7B model, and during its fine-tuning, the QLoRA technique was used to minimize computational requirements. The second model, MicroCoderFIM, is based on the StarCoderBase 1B model and supports code infilling. The individual models were compared based on BLEU, CodeBLEU, ChrF++, and ROUGE-L metrics. Model MicroCoderFIM achieves the best adaptation results to the new task, with over 120% improvement in all measured metrics. The weights of the models along with the new dataset are freely accessible on a public repository.
|
|
|
Automatic Creation of Animated Video based on Textual Story
Kuchař, Josef ; Švec, Tomáš (referee) ; Smrž, Pavel (advisor)
The aim of this work is to link a diffusion model for generating human motion with a diffusion model for generating video. The solution uses current methods for generating video and motion. Video generation is carried out using an image generator equipped with an adapter for temporal consistency. The work introduces a method of connecting both diffusion models using the ControlNet network. The created solution allows for generating video from a simple text description, or a detailed scenario. The program was tested in a user study.
|
|
|
Road Transport Analysis Using Neural Networks
Žárský, Daniel ; Musil, Petr (referee) ; Smrž, Pavel (advisor)
Cílem této bakalářské práce je zjednodušit analýzu silničního provozu, která využívá kamerové záznamy, a to poskutnutím prostředku pro automatickou annotaci scény. Práce popisuje obecné technické pricipy využité v kamerovém systému monitorujícím dopravu a navrhuje postup zpracování dat, získaných metodami počítačového vidění, s cílem automatizovaného nasazení systému. Následné zpracování dat využívá klastrovacích algoritmů pro identifikaci a lokalizaci hlavních směrů pohybu účastníků dopravnícho provozu. Na základě těchto výsledků je scéna automaticky annotována. Anotace scény je použitelná jako základ pozdější detekce anomálií v dopravě v reálném čase.
|
|
|
Creating Advertisement Video Using Neural Models
Taipova, Evgeniya ; Švec, Tomáš (referee) ; Smrž, Pavel (advisor)
The aim of this work is to create a system for the automatic generation of advertising videos based on textual descriptions, which will help users without video production experience save time and money. The work consists of two main parts. The first part uses generative models Stable Diffusion and Stable Video Diffusion for the creation of visual content and GPT-3.5 Turbo for creating scripts for advertising videos. The second part is a web application that allows users to input the necessary information for advertisements and to display the finished videos. This system simplifies and accelerates the process of creating various types of advertisements.
|
|
|
Sentiment Analysis of Czech and Slovak Social Networks and Web Discussions
Slúka, Dušan ; Doležal, Jan (referee) ; Smrž, Pavel (advisor)
This bachelor’s thesis deals with the issue of extraction and analysis of data obtained from social networks to understand public opinion on various social topics. The goal is systematic categorization and interpretation of contents. The problem is solved through a platform for opinion extraction and automatic data classification, which allows the creation of thematic subcategories and sorting into them. The result of the work is a system that analyzes social networks and provides deeper insight into public opinion on social topics. The system enables organizations to better understand the dynamics of online discourse. The benefit of this work is the provision of a new tool for the analysis of social issues, which can serve the academic sphere as well as organizations from practice.
|
|
|
Generating Code from Textual Description of Functionality
Zobal, Ondřej ; Nosko, Svetozár (referee) ; Smrž, Pavel (advisor)
Tato práce se zabývá vývojem rozšíření do editoru Visual Studio Code, které pomůže vývojářům udržet kvalitu kódu jazyka Python 3. Rozšíření poskytuje možnost generování komentářů a docstringů, návrhu nových jmen proměnných. Rozšíření využívá velké jazykové modely Transformer s řídkou pozorností pro zpracování výsledků. Výsledky bohužel nekonkurují současné konkurenci, jakou je například GPT-3.5-turbo.
|
|
|
Automatic Additions and Corrections of Wikidata and Wikipedia Based on Information Extraction
Hložek, Matej ; Otrusina, Lubomír (referee) ; Smrž, Pavel (advisor)
This bachelor's thesis is focused on creation of system for automatic extraction of data from articles in English language from internet encyclopedia site Wikipedia. Depending on class given by text classifier, different types of information are extracted from natural language text and from so called infoboxes of individual articles from Wikipedia. Final product of this system is a knowledge base containing all extracted data and classified type. A notable part of this system is an article extractor that extracts infoboxes and first paragraphs of articles from so called wikidump file.
|
guest ::
