Our research activities in the area of biometrics, first published in the 2018 ERK conference paper [COBISS-ID: 12212564] and later in the extended journal paper of the EEE Transactions on Image Processing (IEEE TIP, IF: 6,079), led to a new AI-based state-of-the-art approach to face super-resolution. The developed approach is applicable to i) various computer vision problems, where low image resolution is problematic, ii) different image processing tasks and applications, iii) mobile apps, iv) entertainment industry, v) consumer electronics and related areas where low-resolution imagery represents an issue. The super-resolution approach was developed as part of the doctoral work of junior researcher Klemen Grm and in 2019 received the Max Snijder Award. The award is given out annually by the European Association for Biometrics (EAB) for the best research work conducted in the scope of a PhD in the broader area of biometrics in Europe. The excellence of the research achievement was also identified by the University of Ljubljana that selected our super-resolution approach among the top-10 research achievements of the university in the calendar year 2019. It also need to be noted that the predecessor to the TIP journal paper, i.e., the 2018 ERK publication, received the best paper award at the Pattern Recognition (PR) section of ERK 2018. The award is given out by the Slovenian Pattern Recognition Society. Based on the super-resolution work we also gave a number of invited talks, including the guest lecture at the University of Hertfordshire in the UK [COBISS-ID 12355924]. The popular Slovenian newspaper Delo published a news article titled “Do razločne slike z digitalnim haluciniranjem” in february 2020 and the IEEE Biometrics Council featured a news item about our approach in their newsletter (https://ieee-biometrics.org/images/pdf/Vol32-Newsletter.pdf). The newsletter aims to inform the global biometric community about the latest achievements in biometrics on the global scale. The RTV1 TV-show “Ugriznimo znanost” also included a section about our achievement in their program [COBISS.SI-ID 12294996].
E.02 International awards
COBISS.SI-ID: 12800852Convolutional neural network (CNN) based approaches are the state of the art in various computer vision tasks including face recognition. Considerable research effort is currently being directed toward further improving CNNs by focusing on model architectures and training techniques. However, studies systematically exploring the strengths and weaknesses of existing deep models for face recognition are still relatively scarce. In this paper, we try to fill this gap and study the effects of different covariates on the verification performance of four recent CNN models using the Labelled Faces in the Wild dataset. Specifically, we investigate the influence of covariates related to image quality and model characteristics, and analyse their impact on the face verification performance of different deep CNN models. Based on comprehensive and rigorous experimentation, we identify the strengths and weaknesses of the deep learning models, and present key areas for potential future research. Our results indicate that high levels of noise, blur, missing pixels, and brightness have a detrimental effect on the verification performance of all models, whereas the impact of contrast changes and compression artefacts is limited. We find that the descriptor-computation strategy and colour information does not have a significant influence on performance. The paper received the 2019 IET Premium Award for the best paper published in the IET Biometrics journal (SCI IF: 2,092) in the last two years. Despite the fact that the paper was published only in 2018 it was the most downloaded and most highly cited paper among all papers published in IET Biometrics in the calendar years 2017 and 2018.
E.02 International awards
COBISS.SI-ID: 11838804Program group member Melita Hajdinjak (co)organized in January 2020 a workshop for educators and teachers of natural science subjects and mathematics. The NA-MA POTI project highlights as one of the fundamental building blocks of natural literacy the ability to recognize, explain, and evaluate the interpretation of natural and technological phenomena, processes, laws, and their interconnection/interdependence in systems. The workshop focuses on interpretations with relevant representations (textual, tabular, graphical) at different ages or educational periods from pre-school education in kindergartens to secondary education in general and vocational high schools.
F.18 Transfer of new know-how to direct users (seminars, fora, conferences)
Program group member, Janez Žiber, developed and implemented (as part of his research activities) a mathematical epidemiology model SIR/SEIR capable of predicting the development of theCOVID-19 epidemic in Slovenia. The model is (among others) used by the expert group of the Slovenian Ministry of Health led by dr. Bojana Beović. Access to simulations of the possible dynamics of the outbreak is via the following web-site: https://apps.lusy.fri.uni-lj.si/appsR/CoronaSim5/. The latest computed models are available from: https://apps.lusy.fri.uni-lj.si/appsR/CoronaSim/ and https://apps.lusy.fri.uni-lj.si/appsR/CoronaSim2/
F.08 Development and manufacture of a prototype
COBISS.SI-ID: 16021763