Institutional Repository of Vilnius University
Optinio simbolių atpažinimo panaudojimas: metodų tyrimas ir pritaikymas /
Abstract
dc:descriptionThis bachelor’s thesis investigated Optical Character Recognition (OCR) methods and developed a Convolutional Neural Network (CNN) model for symbol recognition using the EMNIST Balanced dataset. The work reviewed some of the OCR pipeline stages and focused on optimizing CNN hyperparameters. Experiments demonstrated that careful hyperparameter tuning, including using automated optimization with tools like ”Optuna”, significantly impacts performance. The study highlighted challenges in recognizing visually similar symbols, emphasizing the importance of robust segmentation and feature extraction stages in the OCR process. A comparative analysis indicated that while the results are good, state-of-the-art OCR methods achieve higher accuracies. Future research recommendations include focusing on earlier OCR pipeline stages, employing advanced data augmentation, adaptive segmentation, hybrid architectures, and post-processing techniques to further enhance system robustness and accuracy.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Kozakas, Tomas,
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:ELABAETD239024614&prefLang=en_US