BDAS`05-12 paper searching system
Home > Description
Prof. Michał Kawulok
Institute of Informatics, Silesian University of Technology, Gliwice, Poland
Bio: Michał Kawulok, M.Sc. (2003), Ph.D. (2007), D.Sc. (2015), IEEE member, is an associate professor at the Silesian University of Technology (Gliwice, Poland) and a lead research scientist at Future Processing and KP Labs. He has been involved in numerous successfully completed projects in both academia and industry, and recently he has been leading projects related with super-resolution reconstruction of satellite images, funded by European Space Agency. He has published over 90 papers in peer-reviewed journals and conference proceedings on pattern recognition and image processing. His general research interests are concerned with image processing, pattern recognition and machine learning, with particular attention given to super-resolution reconstruction, face and gesture recognition, linear and non-linear dimensionality reduction techniques and support vector machines.
Title: Reconstructing missing data in digital images: advances and challenges
Abstract: It is not uncommon that information contained in digital images is incomplete, failing to meet the expectations of their users. To mitigate that, a variety of reconstruction strategies have been developed to address different missing data scenarios, concerned with spatial and / or spectral domain. At the beginning of this talk, we will shed light on these problems to better understand the reasons and goals of image reconstruction. This will be followed by outlining recent advancements in this field, mainly attributed to the use of deep convolutional neural networks. In particular, we will focus on super-resolution image reconstruction techniques, whose goal is to address the problem of insufficient spatial resolution, being a noteworthy case of missing information in digital images. Finally, during the talk, we will discuss the biggest challenges in image reconstruction, which the researchers may struggle to overcome in the near future.
Prof. Dr. Dirk Labudde
Bioinformatics group Mittweida (bigM) and Forensic Science Investigation Lab (FoSIL), University of Applied Sciences, Mittweida, Germany
Bio: Dirk Labudde is a professor at the University of Applied Sciences Mittweida, Germany, since September 1, 2009. He received his diploma in 1993 and obtained his Ph.D. in theoretical physics in 1997, both at Rostock University, while also studying medical physics at Kaiserslautern University. He later worked as a lecturer and research assistant at Medical School and Clinical Center for Neurosurgery in Neubrandenburg, Leibnitz Institute for Molecular Pharmacology in Berlin, Technical University Munich, and Technical University Dresden before accepting a professorship position for bioinformatics and forensics at Mittweida.
His main areas of research are algorithms and computational methods in (digital) forensics and structural bioinformatics.
Title: 3D Crime Scene and Disaster Site Reconstruction using Open Source Software
Abstract: Recent developments have given rise to a plethora of soft- and hardware toolkits for three-dimensional reconstruction of objects, locations, places and larger areas. In this respect, reconstruction algorithms have become so efficient that such modeling tasks can be conducted on a single desktop machine in reasonable time. Open source software realizations further provide attractive cost efficient solutions.
In modern forensics, the 3D reconstruction of crime scenes has become popular over the last decade. By integrating temporal data of actions and movements, spatiotemporal models of a given crime can be reconstructed. Such three-dimensional path-time-diagrams can be of great use in ongoing investigations and for archiving as well as reviewing in the future. Additionally, reconstruction procedures can be applied to disaster sites (such as train or plane crash sites, reactor accidents or areas severely harmed by floods, landslides or earthquakes), whereat resulting 3D models can aid in efficient task force planning.
This talk will address the current state of 3D reconstruction processes and spatiotemporal modeling of criminal events by means of open source software. Furthermore, present problems in acquiring underlying data as well as adequate storing are discussed. Requirements for future standards of data quality and processing are elucidated and illustrated on exemplary models obtained from photogrammetric reconstructions.
Prof. Jean-Charles Lamirel
SYNALP team, LORIA, Vandœuvre-lès-Nancy, France
Bio: Dr. Habil. Jean-Charles Lamirel is both lecturer since 1997 and invited Sea-Sky Full Professor at DUT University of Dalian, China, since 2016. He got his PhD in 1995 and his Research Accreditation in 2010. He is currently teaching Information Science and Computer Science at the University of Strasbourg. His main domains of research are Machine Learning Models, Neural Networks, Textual Data Mining, Scientometrics and Social Networks analysis. He is the creator of the paradigms of Data Analysis based on Multiple Viewpoints and Metric based on Feature Maximization. The related models for which it has been proven that they outperform classical models are now used in many challenging Data Mining applications. His work generated an important scientific production: more than 45 invited conferences, PC member of more than 60 international conferences, organizer of 4 international conferences and of more than 18 special sessions in international conferences, editor of special issues in international journals and main author of more than 160 publications in international conferences and journals.
Title: Dealing with the hot topic of urban segregation analysis: new promising combination of approaches
Abstract: With the increasing development of urban megacities, urban segregation analysis and detection is becoming a very important challenge for city organization and urban politics. With more and more fine-grained and massive data becoming available these last years, individual-based models are now made possible in practice. However, this problem needs to exploit complex mathematical tools to be suitably solved. Hence, usual statistical indices have two limitations: they are dependent on the arbitrary definition of local neighborhoods or spatial units, leading to the well-known modifiable areal unit problem, and they are scalar quantities, which means that one single number is supposed to summarize the entire information and complexity in the data. We thus introduce and explore in this talk a very recently introduced mathematical object called multiscalar fingerprint, containing all possible and all scale individual trajectories in a city. We more specifically show that the use of clustering combined with specific metrics for assessing features contributions to clusters allows to explore this new complex object and to single out hotspots of segregation. We illustrate how clustering allows to see where, how and to which extent segregation occurs.
Dr. Paweł Kasprowski
Silesian University of Technology, Gliwice, Poland
Bio: Dr. Pawel Kasprowski is an Assistant Professor at Institute of Informatics, Silesian University of Technology, Poland. He has experience in both eye tracking and data mining. His primary research interest includes using data mining methods to analyze eye movement signal. Dr. Pawel Kasprowski teaches data mining at the University as well as during commercial courses. In the same time, he is an author of numerous publications concerning eye movement analysis.
Title: Introduction to Deep Learning in Keras/Tensorflow
Abstract: Recently deep learning has become a hype word in computer science. Many
problems, which till now could be solved only using sophisticated algorithms,
can be now solved with specially developed neural networks.
The tutorial aims to introduce the basics of Keras/Tensorflow library in Python, and to show how to apply deep learning frameworks in sciencific research.
There is a common belief that to use neural networks a strong mathematical background is necessary as there is much theory which must be understood before starting working. There is also a belief that, because most deep learning frameworks are just libraries in programming languages, it is necessary to be a programmer and have knowledge of the programming language that is used.
While both abilities are beneficial, because they may help in achieving better results, this tutorial aims to prove that deep networks may be used even by people who know only a little about the theory. I will show you ready-to-use networks with exemplary datasets and try to explain the most critical issues which you will have to solve when preparing your own experiments. After the tutorial, you will probably not become an expert in deep learning, but you will know how to use it in practice with your data.