Search  English (United States) Hrvatski (Hrvatska)

innovative promotional partnershipArtificial Intelligence towards EU Multilingualism

Technical co-sponsorship

 
MIPRO 2022 - 45th Jubilee International Convention

SP - MIPRO Junior - Student Papers

Monday, 5/23/2022 3:00 PM - 7:00 PM, Leut I, Hotel Admiral, Opatija

Hybrid Event

Event program
Monday, 5/23/2022 3:00 PM - 7:00 PM,
Leut I, Hotel Admiral, Opatija
3:00 PM - 5:00 PMPapers 
1.O. Kostyleva (Institute of Mathematics and Information Technologies, Irkutsk State University, Irkutsk, Russian Federation), V. Paramonov, A. Shigarov (Matrosov Institute for System Dynamics and Control Theory, Siberian Branch, Russian Academy of Scien, Irkutsk, Russian Federation), V. Vetrova (University of Canterbury, Christchurch, New Zealand)
Towards Comparison of Table Type Taxonomies 
Tables are ubiquitously used to share relational data in various media and formats. Particularly, an enormous number of HTML tables are contained in web pages. They are a valuable data source for applications of web mining, question-answering, and knowledge base construction. However, not all HTML tables are genuine (i.e. containing relational data). Most of them are utilised as means of layout and navigation. In turn, the genuine tables can have different features of layouts, formatting, and content. One of the prevalent problems in web table extraction is to determine main functional and layout types and properties of tables wide-spread on the Web. Currently, a wide range of taxonomies of table types and properties is available to researchers and practitioners. All of these taxonomies could be utilised for table type classification in order to choose further type-specific treatment of tabular data. The existing taxonomies provide similar table types but use the confusing terminology. This paper is an attempt to overview the existing taxonomies of table types by matching their terminology and comparing them qualitatively.
2.D. Beronić, L. Modrić, B. Mihaljević, A. Radovan (RIT Croatia, Zagreb, Croatia)
Comparison of Structured Concurrency Constructs in Java and Kotlin – Coroutines and Virtual Threads 
Ubiquitous multi-core processors with a significant increase in computing power resulted in an omnipresent expansion of concurrent server applications. However, modern multithreaded applications exposed a substantial number of efficiency-related challenges. Consequently, lightweight structured concurrency constructs emerged in various multithreaded applications, as the traditional heavyweight threads model is expensive in regards to memory due to its high dependency on OS kernel threads. Modern programming languages such as Kotlin and Java are both built on the Java Virtual Machine (JVM) and are commonly used in mobile application development and server-side applications, offering, by default, the traditional threads approach. However, Kotlin also includes support for a lightweight concurrency model with coroutines, while Java's virtual threads, announced in the OpenJDK's Project Loom, are still experimental. Such contemporary concurrency implementations are primarily enabling an increase in application performance and efficiency. This paper presents an overview of different approaches to structured concurrency and explores their implementation in programming languages Java and Kotlin. It provides a comparative analysis of traditional threads with coroutines in Kotlin and virtual threads in Java. Based on the conducted testing using benchmarks, we analyzed their performance, described implementation differences, and explored their utilization possibilities and adaptation to real-world use-case scenarios.
3.V. Kopčok, T. Kramberger, R. Kramberger, B. Nožica (Tehničko veleučilište u Zagrebu, Zagreb, Croatia)
Računalno potpomognuto određivanje virusnih varijanti i heterogenosti virusne populacije 
Obrada velikih setova bioloških podataka izuzetno je kompleksna zbog problema povezivanja biologije i računarstva kao dvije odvojene grane istraživanja. Ovaj rad bavi se obradom podataka i otkrivanjem varijanti unutar populacije mikroorganizama virusa. Koriste se unaprijed pripremljeni setovi podataka dobivenih sekvenciranjem genoma nad kojima se vrši pročišćavanje i obrada podataka kako bi se naposljetku izračunala Shannonova entropija i Nukleotidna raznolikost. Rezultati obrade podataka i izračuni Shannonove entropije i Nukleotidne raznolikosti su grafički vizualizirani kako bi se odredila heterogenost virusne populacije. Heterogenost virusne populacije pobliže naznačava mogućnost stabilnih mutacija virusa, što može utjecati na učinkovitost postojećih cjepiva. Rad pojašnjava kako je moguće iz velikog i ljudima nepreglednog seta podataka dobiti vrijedne informacije o mutacijama unutar populacije virusa. Pomoću tehnike objašnjene u radu moguće je obraditi bilo koji set podataka i izvući zaključke prilikom istraživanja virusa. U vrijeme COVID-19 je velika važnost na što ranijem otkrivanju mogućih mutacija virusa i razvoju znanstvenog smjera bioinformatike koji će računalima potpomognuti biološka istraživanja i na taj način ubrzati procese.
4.M. Ilijanic, D. Jaksic, P. Poscic (University of Rijeka, Department of Informatics, Rijeka, Croatia)
Intrusion detection using data mining – an overview of methods and their success 
Problem of processing large volumes of data in a shorter amount of time is a regular occurrence nowadays. This is due to rapidly evolving technologies and Internet being used as the primary source for communication, viewing and searching information, performing transactions, etc. This results in frequent thefts of personal and professional data, as well as an increase of malware or SQL attacks. These are some of the most difficult problems for computers and networks to solve, as well as for information technology security specialists. Numerous tools and methods for detecting and suppressing malicious intrusions are no longer sufficient, so data mining is often being used for that purpose. This paper explains the types of intrusion detection systems and its techniques, the types of intrusions themselves, and briefly describes the most common datasets used in intrusion detection. The definition of data mining and the most common methods and algorithms of data mining are explained. An overview of related work in this field was given, as well as some conclusions based on this analysis. It was concluded that the Random Forest algorithm is the most successful in detecting intrusions, but that the best way to prevent intrusion is to create hybrid models.
5.D. Sladić, T. Kramberger, R. Kramberger, B. Nožica (Tehničko veleučilište u Zagrebu, Zagreb, Croatia)
Određivanje kompleksnosti glazbenog žanra pomoću generativnih modela neuronskih mreža 
Trenutno postoji velik broj različitih žanrova i pod žanrova glazbe. Svaki od žanrova uvelike se razlikuje po kompleksnosti reprodukcije i skladanja. Ovaj rad bavi se generiranjem različitih žanrova glazbe pomoću umjetnih neuronskih mreža s ciljem mjerenja točnosti predviđanja idućeg glazbenog zapisa kako bi se odredila kompleksnost žanra. Koriste se MIDI setovi podataka koji su podijeljeni u različite žanrove glazbe nad kojima se potom vrši treniranje neuronskih mreža. Kako bi se izbjegla mogućnost pristranosti neuronske mreže, koriste se tri različite arhitekture: LSTM, GRU i Transformer. Nad neuronskim mrežama se mjeri točnost svakih 10 epoha treniranja nad testnim setom podataka. Na posljetku se uzima najveća točnost koja pokazuje rezultat predviđanja idućeg tonskog zapisa MIDI datoteke za svaku neuronsku mrežu zasebno. Rezultati mjerenja točnosti predviđanja idućeg zapisa neuronske mreže potvrđuju inicijalnu hipotezu. Žanrovi glazbe uvelike se razlikuju svojom kompleksnošću u vidu predviđanja sljedećeg muzičkog zapisa na temelju prethodnih od strane neuronske mreže. Na taj način moguće je zaključiti da postoji spona između kompleksnosti predviđanja glazbe neuronskom mrežom i žanra glazbe nad kojim je ista trenirana. Pomoću ove tehnike moguće je obraditi bilo koji žanr glazbe i pomoću točnosti dobivene neuronskom mrežom odrediti kompleksnost žanra.
6.L. Matošević, A. Jović (University of Zagreb, Faculty of Electrical Engineering and Computing, Zagreb, Croatia)
Accurate Detection of Dementia from Speech Transcripts using RoBERTa Model 
Dementia is a serious disease that is very common in the elderly population. Automatic detection of dementia is a difficult task that may involve the analysis of acoustic features of speech, linguistic features of transcripts, and mental state exams. In this work, we explore the limits of using speech transcripts from doctor-patient conversations to detect dementia. The dataset is prepared from Pitt corpus, which is a part of DementiaBank, a shared database of multimedia interactions for studying communication in dementia. We use a sophisticated natural language processing approach, namely RoBERTa, which addresses the problem by using transformers and self-attention mechanism. We compare RoBERTa with a baseline BERT model. We show that dementia detection using well-prepared speech transcripts alone can lead to detection rates above 90% for RoBERTa model in a near-balanced dataset, outperforming the baseline model.
7.F. Butić (Fakultet elektrotehnike i računarstva, Zagreb, Croatia), A. Radovan, M. Pajas (BISS Ltd., Zagreb, Croatia), B. Mihaljević (RIT Croatia, Zagreb, Croatia)
Shipping Boxes Feature Extraction on Conveyor Belts Using Real-Time Object Detection Systems 
The use of real-time object detection systems has significantly increased in the previous decade, ranging from personal open-source projects to industry solutions, as is the case with autonomous vehicles. New state-of-the-art models surface on a yearly basis, if not multiple times a year, and stand as a testament to the velocity of the field's rapid advancement. This paper will focus on the use of such a model on objects such as shipping boxes on conveyor belts used in warehouses to improve logistics. The research is based on the YOLO (You Only Look Once) object detection system with low-quality image input. The model can annotate where on the conveyor belt the objects are and give information on features of the objects such as their size, color, and shape. The images have been extracted from surveillance cameras inside shipping warehouses. Significant results with real-time results have been obtained from video cameras that have high-noise frames, which means that better camera placement and higher quality images can only improve existing results and model performance.
8.M. Barišić, A. Jović (University of Zagreb, Faculty of Electrical Engineering and Computing, Zagreb, Croatia)
Cardiac Arrhythmia Classification from 12-lead Electrocardiogram Using a Combination of Deep Learning Approaches 
Traditionally, electrocardiogram (ECG) signals are recorded and monitored over a period of time and finally analyzed by an expert. Automatic classification of cardiac arrhythmias has the potential to improve diagnostics. In this work, we explore the use of representation learning from ECG signals for cardiac arrhythmia classification. The dataset was created from the CPSC, CPSC-Extra, and The Georgia 12-lead ECG Challenge (G12EC) databases. We use a sophisticated deep learning approach for representation learning and classification, namely a combination of Generative Adversarial Network (GAN), a Convolutional Auto-Encoder (CAE), and a Long Short Memory (LSTM) classifier. CAE was used to compress the input signal that serves as input to the LSTM classifier. We implemented GANs to balance the distribution of data, in which the analyzed heartbeats belong to an arrhythmia type. The high classification results show that the use of the complex deep learning approach is suitable for solving the problem.
5:00 PM - 5:15 PMBreak 
5:15 PM - 7:00 PMPapers 
1.D. Beronić, N. Novosel, B. Mihaljević, A. Radovan (RIT Croatia, Zagreb, Croatia)
Assessing Contemporary Automated Memory Management in Java – Garbage First, Shenandoah, and Z Garbage Collectors Comparison 
Selecting the appropriate automated memory management approach directly impacts application performance and is considered one of the crucial factors in contemporary memory management. In Java, applications are commonly executed within the Java Virtual Machine (JVM), and, for years, it has offered several garbage collection techniques to choose from. The Garbage First (G1) garbage collector (GC) became the standard in 2017 and is currently a widely used garbage collection algorithm. Other GCs evolved in the meantime, such as Shenandoah GC and Z Garbage Collector (ZGC), and they were recently promoted from experimental features to production-ready. Since currently, there are several standard GCs to choose from, we believe that the performance distinctions they manifest should be more thoroughly examined with experimental benchmarks and real-life use cases. Given the importance of GC in automated memory management performance, it is worthwhile to investigate how those GC algorithms handle various memory issues and perform in regards to heap allocation, CPU usage, and time consumption. This paper focuses on the comparative analysis of the default G1 and the Shenandoah and Z GCs, and compares the number of measures in selected application tests from the popular benchmarking suites with all three standard GCs available.
2.P. Lončar (University of Split, Faculty of Science, Split, Croatia)
Internet of Musical Things and Music Data Visualization 
Data visualization is one of the significant disciplines of data science. This paper presents an overview of the growing research field of the Internet of Musical Things (IoMusT) and the visual aspect of music. The aim is to analyze the relation and influence of information technology in the music domain, music creation, and music development. The author provides a brief overview of relevant studies in evaluating the usage of various music interactive interfaces. Analysis of different perspectives of music visualization, including visualization of music elements, melodies, musical instruments, discography, and the comparison of music styles are presented. The paper gives a data analysis from one of the leading streaming applications, Spotify, correlating with songs on Billboard charts from the 2010s. The results of conducted data analysis are presented graphically. The musical features of several selected popular songs are visualized. Visualization is based on the MIDI format, a format important in creating digital music. This analysis emphasizes the importance of new concepts and music data visualization contribution to limitless enrichment of musical possibilities and music content.
3.J. Dobruna (University of Ljubljana, Ljubljana, Slovenia), E. Spahiu (University of Prishtina, Prishtina, Kosovo), M. Pogačnik , M. Volk (University of Ljubljana, Ljubljana, Slovenia)
5G Streaming: IP based vs. High-Power High-tower broadcast 
The new generation of the cellular communication (5G) presents a highly flexible and scalable network that is designed to connect virtually everyone and everything including machines, objects, and devices. 5G will provide higher bandwidth, lower end to end delays and improved reliability. Traffic in 5G communication networks will be dominated by video applications such as mobile broadcasting, augmented reality and ultra-high content quality delivery. This new generation is expected to handle ultra-high-definition video streaming and provide multimedia services in an efficient way in order to meet the users’ expectation. In particular, 5G Broadcast has a wide application in facilitating the distribution of audio-visual media content, when covering popular live events with large audience. Broadcasting in 5G can be accomplished using either a high-power high-tower (HPHT) transmitter or by IP-based streaming. In this paper, we evaluated the Quality of Experience (QoE) for different broadcasting configurations. The different scenarios are simulated through MATLAB. From simulations we can conclude that deployment of HPHT base stations will reduce delays and packet loss rate, whereas QoE is improved.
4.K. Kacperowski, E. Szymańska, M. Purzycki, M. Skrok, M. Pietrzak (Wroclaw University of Science and Technology, Wroclaw, Poland)
Optimized Laser Triangulator for Underwater Robot Vision 
The underwater world holds plenty of undiscovered secrets, not yet fathomed due to the hostile conditions in the ocean depths. One of the problems that can be encountered in such an environment is limited or lack of visibility. Average cameras, even when equipped with additional lighting systems, do not allow for data collection from a greater distance. However, a different kind of imaging can be used. Based on the laser triangulation approach, a three-dimensional model of the surroundings can be created and subsequently analysed. There are numerous devices that use this method and similar technologies such as light detection and ranging, but they are characterised by high complexity and therefore price. In this paper, the design of a device intended to be mounted on AUVs and ROVs is presented. The operation principles for the device are simplified, allowing for relatively cheap creation, yet holding the required features. This research is complemented with tests in above- and underwater environments, measuring the surrounding elements’ dimensions as well as distances to them.
5.M. Purzycki, A. Komorowska, A. Ilnicka, J. Papież, E. Szymańska (Wroclaw University of Science and Technology, Wroclaw, Poland)
From ROVs to AUVs - Optimization and Analysis of Underwater Vehicles Design 
Underwater vehicles are capable of relieving humans of task execution in hazardous environments, such as seas or oceans. Elevated pressure, poor visibility and the importance of structural tightness are just a few factors that must be considered during the design and later in the implementation phase. In this paper, a comparative analysis of the underwater robot design steps is shown using the example of the PWr Diving Crew project, where simplified Remotely Operated Vehicles (ROVs) were improved to Autonomous Underwater Vehicles (AUVs) with enhanced environment perception and capability of autonomous task performance. Robots’ characteristics and the altered approach to the problem in successive vehicle generations are described, and the conclusions on the structure optimization are drawn based on the experience gained in the process.
6.D. Škrlec, V. Šimović, A. Penđer (Zagreb University of Applied Sciences, Zagreb, Croatia)
Arduino Based Quadcopter 
Aircraft are increasingly being used for all sorts of tasks, of which drone video shooting is the most famous example. In order to reduce the cost of buying an aircraft, there is the possibility of making your own aircraft. The aim of this work was to make an aircraft operated by the Arduino as the main "brain" instead of a commercial flight controller. The successful and rather cheap construction of the Arduino spacecraft proved that everyone can have their own drone with a little effort.
7.P. Mage, K. Krajček Nikolić (Fakultet prometnih znanosti, Zagreb, Croatia)
Modeling the Effect of One Engine Inoperative for Bada Flight Performance Model  
The paper deals with the augmentation of the existing BADA 3 performance model to incorporate one engine inoperative flight performance option. BADA aircraft performance model was made by EUROCONTROL (European Organization for the Safety of Air Navigation) and represents a base of data relevant for aircraft performance. The model is mostly used for simulating aircraft trajectories in airspace where the behaviour of aircraft is realistic if it flies ordinary regimes, but less realistic when it comes to engine failures. Flying with an inoperative engine is specific and differs from ordinary operations in terms of operational limits that affect operational efficiency and safety itself. The modelling principle was tested using available data from the aircraft operation manual (AOM) and aircraft flight manual (AFM) for twin-engine turboprop passenger aircraft Dash 8 Q400 used for regional transport. The objective is to analyse and compare changes in aircraft performance after an engine failure during flight. Corrections for the current aerodynamic model are given for the case of cruise flight with one engine inoperative. Results of modelling are commented and recommendations for further research are given.
8.A. Šimec, J. Kostanjevac, L. Tepeš Golubić ( Zagreb University of Applied Sciences, Zagreb, Croatia)
Neural Networks and Their Application for Image Recognition 
This paper will explain what an artificial neural network is and how it works in theory. To make it easier to explain how an artificial neural network works, the first part will explain the terminology, the idea behind neural networks and their role in machine learning. In the continuation of the paper, the development of neural networks will be presented, as well as the principle of operation of the most important models that will be used through examples, and special attention will have focus to explain the concepts of neural networks. Through various applications in the examples, we will analyze the architecture of neural networks on image recognition using web applications and explain the concepts of convolutional neural networks.

Basic information:
Chairs:

Benjamin Kušen (Croatia), Dubravko Sabolić (Croatia), Lovro Božičević (Croatia)

Registration / Fees:
REGISTRATION / FEES
Price in EUR
EARLY BIRD
Up to 9 May 2022
REGULAR
From 10 May 2022
Members of MIPRO and IEEE
230
260
Students (undergraduate and graduate), primary and secondary school teachers
120
140
Others
250
280

The discount doesn't apply to PhD students.

Contact:

Benjamin Kušen
KSET
Unska 3
HR-10000 Zagreb, Croatia

GSM: +385 98 985 0111
E-mail: benjamin.kusen@fer.hr

The best papers will get a special award.
Accepted papers will be published in the ISSN registered conference proceedings. Presented papers in English will be submitted for inclusion in the IEEE Xplore Digital Library.
.............
There is a possibility that the selected scientific papers with some further modification and refinement are being published in the following journals: Journal of Computing and Information Technology (CIT)MDPI Applied ScienceMDPI Information JournalFrontiers and EAI Endorsed Transaction on Scalable Information Systems.


Location:

Opatija is the leading seaside resort of the Eastern Adriatic and one of the most famous tourist destinations on the Mediterranean. With its aristocratic architecture and style, Opatija has been attracting artists, kings, politicians, scientists, sportsmen, as well as business people, bankers and managers for more than 170 years.

The tourist offer in Opatija includes a vast number of hotels, excellent restaurants, entertainment venues, art festivals, superb modern and classical music concerts, beaches and swimming pools – this city satisfies all wishes and demands.

Opatija, the Queen of the Adriatic, is also one of the most prominent congress cities in the Mediterranean, particularly important for its ICT conventions, one of which is MIPRO, which has been held in Opatija since 1979, and has attracted more than a thousand participants from over forty countries. These conventions promote Opatija as one of the most desirable technological, business, educational and scientific centers in South-eastern Europe and the European Union in general.


For more details, please visit www.opatija.hr and visitopatija.com.

Download
 
News about event
Currently there are no news
 
Patrons - random
Sveučilište u RijeciFER ZagrebPomorski fakultet RijekaTehnički fakultet RijekaFOI Varaždin