DOTSOFT SA participates in the implementation of the research project entitled "Cultural Dimensions of
Deep Learning", with the acronym "CULDILE", code T1EDK-03785 within the framework of the Single
Action of State Support for Research, Technological Development & Innovation "RESEARCH - CREATE -
INNOVATE" of E.P. "Competitiveness, Entrepreneurship and Innovation (EPANEK)", ESPA 2014 - 2020.
The project is co-financed by the European Union and the European Regional Development Fund.
Organizations cooperating for the implementation of the project:
- P. TSOUGRIANIS AND CO. (HONEST PARTNERS)
- NATIONAL RESEARCH CENTER FOR NATURAL SCIENCES
- DEMOKRITOS EKEFE "D"
- Piraeus PIOP Group Cultural Foundation
- DOTSOFT SA
In this project, an integrated, expert software system for capturing, optimizing, and understanding
culture evidence will be developed. Capturing and pre-processing the image as it comes from a low-cost
scanner is the first stage of processing documents with the aim of improving their quality and creating
an excellent digital copy. Improving the quality of the document, in addition, is a particularly important
stage because it helps the success of the subsequent stages of processing, such as segmentation and
identification, and is necessary due to the particular problems of quality and appearance of the
historical documents to be processed.
Recognition and segmentation are key stages of understanding and lead to the expert assistance of
documentation. It will be automated to indicate and identify repetitive elements such as: initials,
dedication note, typographic ornaments, titles, thumbnails, character families, titles, tables of contents
and indexes, etc., as well as the existence of specific keywords that characterize the content of the
documents or part of these e.g., in a Gospel, its beginning is "that time". In this way, a 'unique' in the
world application (software) of assisting the scientific documentation of culture evidence will be
In the first part (capturing) emphasis will be placed on the use of affordable scanning devices, combined
with development/evolution of Honest Partners' existing HDOC+ document image processing platform.
In the field of assisted documentation, we focus on developing state-of-the-art deep neural networks
methodologies to create expert user assistance and visually assist the user to have the optimal and
fastest simultaneous visual experience of the document pages and of the corresponding documentation
and notes (web based). System outputs (images, documentation) will be imported and managed
through open-source repositories.