Nowoczesne metody automatycznej eksploracji źródeł masowych
Modern methods of automatic exploration of mass sources

Author(s): Krzysztof Kotowski, Zygfryd Wieszok, Maciej Wojsyk
Subject(s): History
Published by: Wydawnictwa Uniwersytetu Warszawskiego
Keywords: computer mass source exploration; computer image analysis; machine learning; Adaptive Vision Studio; automatic text recognition; Google Tesseract
Summary/Abstract: The text presents the capabilities of modern computer image analysis and machine learning methods in tasks related to the exploration of mass sources. The first part briefly discusses ways of digitizing data, the basics of machine learning and the specification of the Adaptive Vision Studio tool with Deep Learning Add-on. Furthermore, a proprietary tool for supporting manual image analysis was proposed. The main part is the case study of measurement cards of soldiers from years 1921–1935. It presents examples of effective automatic segmentation and location of elements like of the image as underlining or stamps. The recognition of machine text using the Google Tesseract library was shown to be equally effective. However, automatic recognition of handwritten text was recognized as an open problem, usually requiring a dedicated solution. At the end, current trends in the field of computer-aided mass source exploration are presented.

• Details
• Contents