It can move autonomously in an unfamiliar confined environment and create a detailed map. This is the role of a six-legged rescue robot whose “brain” was designed and tested by Master’s student Jan Bayer from the Faculty of Electrical Engineering at the Czech Technical University (CTU) in Prague. His precisely executed Master’s project with great potential for society at large won him first place in the prestigious IT SPY competition that names the best IT Master’s theses from Czech and Slovak universities each year. Second place went to Slovakia, to Michal Hucko from the Faculty of Informatics and Information Technologies at the Slovak University of Technology (STU) in Bratislava, whose algorithm helps internet developers identify lost and confused website users in real time. The third-best project, belonging to Dušan Drevický from the Faculty of Information Technology at the Brno University of Technology (VUT), focuses on uncertainty in machine learning models. It helps computers notate x-ray images where the computer isn’t sure if it evaluated them correctly. This year, 17 academics from Czech and Slovak IT Faculties made their selections from nearly 1,600 theses.
The winning project by Jan Bayer from CTU in Prague took the grand prize by an almost unanimous vote. In addition to getting first place, he also won the award from Valeo, the main partner of the competition. The thesis solves the fundamental problem of exploring unfamiliar environments during rescue missions resulting from natural disasters, industrial accidents, and mine collapses, as well as when searching for people in cave systems. The Bayer-designed program that manages the robot can use the information it gathers to come up with a way to map the outlined area as best as possible and determine the navigability of the terrain. This can help rescuers decide whether it is safe to enter the area.
“The system was developed for a six-legged walking robot, which was deployed in various scenarios in both indoor and outdoor environments. The solution was also deployed in the elite US DARPA Subterranean Challenge on a wheeled crawler robot with various sensory equipment, which is used to test solutions for NASA GPL, MIT, and CMU laboratories. And, of course, I’ve included the results in my thesis,” mentions Jan Bayer. “We will now deploy the system, for example, in an urban subterranean exploration scenario on a simulated rescue mission.”
“We were particularly impressed by the fact that the student presented a comprehensive solution that was even able to pass the test used by the development departments of the US civil defense and military. The society-wide contribution and professionalism of such a solution were also recognized by the jurors,” states Professor Mária Bieliková from the Faculty of Informatics and Information Technologies at STU in Bratislava, the academic guarantor of this year’s competition.
Second place went to Michal Hucko from STU in Bratislava. His algorithm, which can be used to personalize websites and applications, could fundamentally improve ergonomics and user-experience projects. From observing user behavior, it can determine, in real time, the moment when the user is probably lost on the website or in the application and respond—for example, by offering a help window or connecting users with online support. It also provides feedback for developers and identifies critical points in the navigation.
“I was looking for a solution that would be comparable to eye-tracking technology. Eye-tracking technology is fairly accurate, but very expensive and requires special components that make it impossible to control in the course of normal contact with the user. Our approach is very similar to biometric authentication, and simply tracking cursor movement produces just as accurate results as costly eye-tracking,” explains Michal Hucko. “The solution is ready for commercial deployment and has already been used live.”
“Given the maturity of Czech and Slovak e-commerce, I consider a solution to this problem a very interesting idea for which real demand exists here among developers. It could help effectively solve problems, for example, in digital onboarding for banks and other services. The author of the thesis also has a clear vision of how to further make use of his project and already has the solution integrated into a functioning product,” says Tomáš Krátký from Profinit, the professional guarantor and co-organizer of the competition from the corporate sphere.
Third place went to Dušan Drevický from Brno VUT. His Master’s thesis responds to the growing role of neural networks and machine learning mechanisms. They often have to make their own decisions and adopt conclusions that they further work with. However, sometimes they mistakenly declare an erroneous statement of their own correct. So, Drevický created and tested an algorithm that when examining x-ray images of teeth can determine for itself how confident it is in its conclusions. This significantly accelerates the planning of dental intervention and eliminates the potential risk of error resulting from computer analysis of the images.
“Today’s widespread use of machine learning models in medicine is still hampered by a lack of information about whether the conclusions that the algorithm accepts are correct. Mistakes in medicine are expensive and serious, but in many cases machine learning models are becoming indispensable—they are better and more accurate than doctors. They also save doctors work and allow them to focus more on tasks that require expert decisions,” explains Dušan Drevický. “My work solves this problem and, as has been proven, significantly increases doctors’ efficiency and their confidence in the use of artificial intelligence in their medical practice. For dentists, which we don’t have enough of, it makes sense in this regard.”
“The role of automatic learning and computer assistance grows each year. It is becoming increasingly clear that correction mechanisms must be developed alongside learning mechanisms. The thesis took on the task of implementing them in an environment where there is no room for error, in medicine, and it has produced very convincing results,” concludes Bieliková.
This year, the People’s Choice Award went to Matúš Mrázik from the Faculty of Management Science and Informatics at the University of Žilina with a program that, when learning from a series of materials, can exclude those that seem to be wrong. So, it avoids introducing errors resulting from poorly presented patterns into its learning mechanisms.
Academics agree that the quality of Czech and Slovak IT students is in no way lagging behind the rest of the world. Their projects take note of the most important global trends and most of them have the chance to break through in global competition.
“This year we assessed a record number of theses that work with or use methods of artificial intelligence, one of the key elements of future technological development. The growing international response to student Master’s projects—for example, the opportunity to present them at scientific conferences abroad—is also pleasing,” concludes Bieliková.
Leoš Dvořák, general manager of the research and development center at Valeo, sees it similarly. According to him, today’s Czech and Slovak Master’s students have the chance to collaborate on projects that will change the face of a wide range of industries—for example, the important Czech and Slovak automotive industry.
“As the world leader in automotive development, we can say that Czechs and Slovaks have truly left their mark on the development of, for example, autonomous vehicles. Indeed, the winning project, which introduced the algorithm of an autonomously controlled robot in an unfamiliar environment, is proof of how advanced the students are in this field.”
Tomáš Krátký from Profinit agrees that collaboration between companies and graduates is highly desirable. Profinit co-organizes the annual IT SPY competition and works closely with universities.
“At Profinit, we have, for example, created our own system of internal education for new colleagues that smoothly follows what they learned at university. Thus, we can brilliantly continue the good work started by the Czech and Slovak universities and maximally evaluate the abilities of former students. And because of this, a wide range of quality Master’s projects will reach the world.”
The tenth annual IT SPY competition evaluated the theses of 1,570 graduates from the 2018/19 academic year from 17 IT faculties across the Czech Republic and Slovakia. The winner, in addition to academic recognition, gets 1,000 EUR to further develop the project.
You can take a look at all the projects at https://www.itspy.cz/en/galerie-nejlepsich/.
Complete competition results:
Grand Prize
Jan Bayer, Faculty of Electrical Engineering, Czech Technical University in Prague
Autonomous exploration of rough terrain with a six-legged walking robot
The winning thesis that deals with the mapping of unfamiliar or altered terrain is essential for rescue missions in times of natural disasters or mine collapses. Not only does it map the given site, it also determines its navigability and verifies if it is safe for rescuers to enter. The newly developed system, deployed on a robot, helps with this. Using the information obtained, it can propose another route in order to be able to map the outlined area in the best way possible. It was also tested on a six-legged wheeled crawler robot, both inside a building and outdoors in rough terrain. The proposed solution was also deployed in several other experiments, including indoor areas and mine corridors that were successfully explored in the DARPA Subterranean Challenge.
Second Place
Michal Hucko, Faculty of Informatics and Information Technologies, Slovak University of Technology in Bratislava
Identification of user confusion in a web application
This thesis deals with the optimization of websites and applications so that artificial intelligence can determine when the user is lost and does not know where to go. The model tracks the movement of the mouse to learn how to recognize user confusion and can offer help based on the user’s previous activity. It can thus guess whether the user is having trouble finding some information and recommend a solution to the problem. Furthermore, it also serves developers and website owners, as tracking user confusion can help them improve the layout and optimization of the site. This solution was tested on the travel web portal FiroTour and has been integrated into the portal YesElf.
Third Place
Dušan Drevický, Faculty of Information Technology, University of Technology in Brno
Deep learning model uncertainty in medical image analysis
Dušan Drevický applied neural networks to automatically locate key points in x-ray images of skulls. Thanks to his improvements, the neural network can evaluate whether it is certain of its conclusions or not. Based on this information, the doctor knows if she needs to process the image manually. This innovation significantly accelerates the planning of dental intervention. The thesis was created in cooperation with TESCAN 3DIM and will be incorporated into the medical diagnostic software they develop.
Fourth–Eighth Place (in alphabetical order)
Šimon Lomič, Faculty of Information Technology, Czech Technical University in Prague
Taking and breaking games
This thesis approaches game solutions via combinatorial game theory. The thesis introduces analytical solutions for new and previously investigated variants of these games and proves that some variants almost certainly do not have an easy solution.
Matúš Mrázik, Faculty of Management Science and Informatics, University of Žilina
Reversible feature extraction
This thesis represents a shift in machine learning, where the program created can evaluate what doesn’t belong to the preset pattern and thus can make manual checks of feedback control patterns quicker and easier.
Vojtěch Polášek, Faculty of Informatics, Masaryk University in Brno
Argon2 security margin for disk encryption passwords
This thesis deals with cybersecurity and verifies the actual effectiveness of the Argon2 encryption key.
Jakub Strmeň, Faculty of Natural Sciences, Matej Bel University in Banská Bystrica
Parallel computer algebra system for structural analysis of snarks
The thesis presents important findings about the properties of graphs known as snarks, which can contribute to solving problems with graph coloring and thus also the optimization of production processes and network communication.
Martin Süss, Faculty of Business and Economics, Mendel University in Brno
Recognizing named entities in text
It can improve machine reading and automatic recognition of the significance of the text.
VALEO prize for the best Master’s thesis in the field of computer vision and sensory data processing The prize awarded by competition partner Valeo for the project that offers the greatest benefit to the field of enterprise systems.
Jan Bayer, Faculty of Electrical Engineering, Czech Technical University in Prague
Autonomous exploration of rough terrain with a six-legged walking robot
People’s Choice
The prize awarded on the basis of online votes from the public
Matúš Mrázik, Faculty of Management Science and Informatics, University of Žilina
Reversible feature extraction