University of Rostock


University of Rostock – Alma Mater Rostochiensis – Traditio et Innovatio: Founded in 1419, the University of Rostock is the oldest in the Baltic Sea Region. True to the motto “Traditio et Innovatio”, the University of Rostock has constantly further developed. The multitude of new buildings represents the university’s modernity. Today, with about 14,000 students and 2,933 staff members, the University of Rostock offers fascinating perspectives into nearly all scientific fields. With the four profile lines Life, Light and Matter / Maritime Systems / Aging Science and Humanities / Knowledge – Culture – Transformation, the University of Rostock has at its disposal excellent interdisciplinary research fields in the areas of natural and technological sciences, medicine, life sciences, humanities and cultural studies.

The Computational Intelligence Technology Lab (CITlab) Project Group of Prof. Dr. Roger Labahn is part of the Mathematical Optimization Group (head: Prof. Dr. Konrad Engel) at the Institute of Mathematics of the University of Rostock. As its host, the Institute of Mathematics strongly supports CITlab’s work by ensuring all necessary working conditions, thus maintaining one of its most important application related actions. The group’s scientific focus is on algorithms and their mathematical foundations, where the institute environment allows incorporating application-oriented knowledge and mathematical support from different areas like Optimization, Stochastic & Mathematical Statistics and Numerical Mathematics – each one covering important issues at CITlab’s research and technology groundings.

The CITlab group has several years of experience in the development of application-oriented algorithms and engines for text and handwriting recognition. Its work explicitly did extend towards both research and innovation as well as to technology development and implementation. One of the essential backgrounds is the tight and long-term research and development collaboration with PLANET: This is an SME realizing the technology and software development and the industry standard implementation part. For 20 years now, PLANET successfully uses computational intelligence and neural network based technology for real market applications, today including the field of document recognition and understanding.

As for academic test scenarios, both CITlab and the Neural Network based recognition technology did prove their abilities by successfully participating in international competitions, mainly hosted by two world-wide leading conference series, the International Conference on Document Analysis and Recognition (ICDAR) and the International Conference on Frontiers in Handwriting Recognition (ICFHR): Rank 3 at the NIST Open Handwriting Recognition and Translation Evaluation OpenHaRT 2013, first ranks at the two Handwriting Text Recognition on a transScriptorium Dataset competitions HTRtS 2014 & 2015, five first ranks at the ANCESTRY Word Recognition from Segmented Historical Documents ANWRESH 2014, and Ranks 1 at the Keyword Spotting for Handwritten Documents competition KWS 2015 as well as at the Handwritten Scanned Document Retrieval Task ImageCLEF 2016. Various of these results showed remarkable advances of CITlab’s technology over the other contributions.

Since 2008, the CITlab group has continuously been working in application-oriented projects funded by the German provincial or federal governments, respectively, via the SME partner PLANET, it was closely associated to the FP7 project ORGANIC, and since 2016, CITlab is one of the leading technology partners in the Horizon-2020 project READ. Moreover, CITlab is running another Machine Learning related project in collaboration with the Max-Planck-Institute for Plasmaphysics based on the EUROfusion programme of the EU.

Currently, the CITlab group consists of 7 academic researchers, and moreover, it has continuous technology and development support from the SME PLANET regarding industry standard software implementation.

All this enables the CITlab project group to successfully contribute to the NewsEye project’s research and technology tasks related to and being necessary for Text Recognition and Article Separation: innovative improvements of established methods, and investigating new research grounds for Layout Analysis, Automatic Text Recognition and Article Separation based on state-of-the-art neural methods of Machine Learning domain.