Institutional Repository of Vilnius University
Lentelių struktūrų aptikimas ir informacijos nuskaitymas buhalteriniuose dokumentuose /
Abstract
dc:descriptionOften times the need arises to digitize tabular data found in financial and other documents. Typing everything by hand can be a time consuming and expensive process therefore many tools are being made to automate this task. In this work two such table structure recognition algorithms were compared: “TableNet” and “DeepTabStR”. A deep neural network from the “TableNet” approach was implemented and evaluated with a new financial document dataset. In order to get better results for financial documents further improvements to the network were made: additional image preprocessing using the “OpenCV” library, adding financial documents to the training data and introducing deformable convolution layers to the network. All three improvements resulted in better image segmentation.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2021
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Kybartas, Lukas,
- Contributors dc:contributor
-
- Valaitis, Vytautas
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD107116097&prefLang=en_US