{"id":{"repo_id":"vilnius","oai_identifier":"oai:vu.lt:elaba:308337699"},"canonical_url":"https://search.ndltd.org/etd/vilnius/oai:vu.lt:elaba:308337699","repository":{"repo_id":"vilnius","name":"Vilnius University","base_url":"https://epublications.vu.lt/oai"},"display":{"title":"Ranka rašyto lietuviško teksto atpažinimas /","abstract":"This bachelor thesis explores the issue of handwritten text recognition in Lithuanian (HTR), emphasizing the influence of synthetic data and different training approaches on the model's recognition accuracy. The theoretical part reviews the main principles of HTR and synthetic data generation, as well as the application of CNN, BiLSTM and CTC components in HTR systems. The experimental part of the study involved gathering an original corpus of handwritten Lithuanian sentences from specially designed forms and developing an automatic pre-processing procedure for image perspective correction, sentence segmentation and annotation generation. Synthetic data consisted of sentences written using hand-like fonts and supplemented by image augmentation. CRNN models were trained using different real-to-synthetic data ratios under standard and progressive training methods. In the progressive approach, the synthetic data subset was refreshed after each training cycle, while a separate holdout dataset was used to evaluate the model‘s generalization ability. The experiments showed that only a combination of real and synthetic data could guarantee satisfactory recognition accuracy; the progressively trained model achieved a holdout character error rate of 0.1900 using 1:1 real-to-synthetic data ratio. Practical tests also showed promising sentence recognition performance. The obtained results may serve as a basis for further research on handwritten text recognition for Lithuanian and other less studied languages.","abstract_html":"This bachelor thesis explores the issue of handwritten text recognition in Lithuanian (HTR), emphasizing the influence of synthetic data and different training approaches on the model&#x27;s recognition accuracy. The theoretical part reviews the main principles of HTR and synthetic data generation, as well as the application of CNN, BiLSTM and CTC components in HTR systems. The experimental part of the study involved gathering an original corpus of handwritten Lithuanian sentences from specially designed forms and developing an automatic pre-processing procedure for image perspective correction, sentence segmentation and annotation generation. Synthetic data consisted of sentences written using hand-like fonts and supplemented by image augmentation. CRNN models were trained using different real-to-synthetic data ratios under standard and progressive training methods. In the progressive approach, the synthetic data subset was refreshed after each training cycle, while a separate holdout dataset was used to evaluate the model‘s generalization ability. The experiments showed that only a combination of real and synthetic data could guarantee satisfactory recognition accuracy; the progressively trained model achieved a holdout character error rate of 0.1900 using 1:1 real-to-synthetic data ratio. Practical tests also showed promising sentence recognition performance. The obtained results may serve as a basis for further research on handwritten text recognition for Lithuanian and other less studied languages.","abstract_has_math":false,"creators":["Noreika, Normantas,"],"institution":"Institutional Repository of Vilnius University","degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2026,"date_issued":"2026","date_published":"2026","updated_at":"2026-11-19T10:12:59Z","subjects":[],"languages":["lit"],"rights":["info:eu-repo/semantics/openAccess"],"rights_urls":[],"identifier_entries":[],"person_identifier_entries":[]},"links":{"outbound_url":"https://repository.vu.lt/VU:ELABAETD308337699&prefLang=en_US","outbound_label":"Repository record","outbound_source":"dc:identifier","record_url":"https://repository.vu.lt/VU:ELABAETD308337699&prefLang=en_US","record_label":"Repository record","record_kind":"record","record_source":"dc:identifier","download_url":null,"download_label":null,"download_source":null,"access_status":"open","embargo_until":null,"access_note":null},"source_record":{"url":"https://epublications.vu.lt/oai?verb=GetRecord&metadataPrefix=oai_dc&identifier=oai%3Avu.lt%3Aelaba%3A308337699","prefix":"oai_dc"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:creator","label":"Author","values":["Noreika, Normantas,"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2026"]},{"key":"dc:publisher","label":"Institution","values":["Institutional Repository of Vilnius University"]},{"key":"dc:relation","label":"Dc Relation","values":["https://epublications.vu.lt/object/elaba:308337699/308337699.pdf"]},{"key":"dc:type","label":"Dc Type","values":["info:eu-repo/semantics/bachelorThesis"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["lit"]},{"key":"dc:rights","label":"Dc Rights","values":["info:eu-repo/semantics/openAccess"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["https://repository.vu.lt/VU:ELABAETD308337699&prefLang=en_US"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["This bachelor thesis explores the issue of handwritten text recognition in Lithuanian (HTR), emphasizing the influence of synthetic data and different training approaches on the model's recognition accuracy. The theoretical part reviews the main principles of HTR and synthetic data generation, as well as the application of CNN, BiLSTM and CTC components in HTR systems. The experimental part of the study involved gathering an original corpus of handwritten Lithuanian sentences from specially designed forms and developing an automatic pre-processing procedure for image perspective correction, sentence segmentation and annotation generation. Synthetic data consisted of sentences written using hand-like fonts and supplemented by image augmentation. CRNN models were trained using different real-to-synthetic data ratios under standard and progressive training methods. In the progressive approach, the synthetic data subset was refreshed after each training cycle, while a separate holdout dataset was used to evaluate the model‘s generalization ability. The experiments showed that only a combination of real and synthetic data could guarantee satisfactory recognition accuracy; the progressively trained model achieved a holdout character error rate of 0.1900 using 1:1 real-to-synthetic data ratio. Practical tests also showed promising sentence recognition performance. The obtained results may serve as a basis for further research on handwritten text recognition for Lithuanian and other less studied languages."]},{"key":"dc:format","label":"Dc Format","values":["application/pdf"]},{"key":"dc:title","label":"Title","values":["Ranka rašyto lietuviško teksto atpažinimas /","Handwritten lithuanian text recognition."]}]}],"canonical_facts":{"dc:creator":["Noreika, Normantas,"],"dc:date":["2026"],"dc:description":["This bachelor thesis explores the issue of handwritten text recognition in Lithuanian (HTR), emphasizing the influence of synthetic data and different training approaches on the model's recognition accuracy. The theoretical part reviews the main principles of HTR and synthetic data generation, as well as the application of CNN, BiLSTM and CTC components in HTR systems. The experimental part of the study involved gathering an original corpus of handwritten Lithuanian sentences from specially designed forms and developing an automatic pre-processing procedure for image perspective correction, sentence segmentation and annotation generation. Synthetic data consisted of sentences written using hand-like fonts and supplemented by image augmentation. CRNN models were trained using different real-to-synthetic data ratios under standard and progressive training methods. In the progressive approach, the synthetic data subset was refreshed after each training cycle, while a separate holdout dataset was used to evaluate the model‘s generalization ability. The experiments showed that only a combination of real and synthetic data could guarantee satisfactory recognition accuracy; the progressively trained model achieved a holdout character error rate of 0.1900 using 1:1 real-to-synthetic data ratio. Practical tests also showed promising sentence recognition performance. The obtained results may serve as a basis for further research on handwritten text recognition for Lithuanian and other less studied languages."],"dc:format":["application/pdf"],"dc:identifier":["https://repository.vu.lt/VU:ELABAETD308337699&prefLang=en_US"],"dc:language":["lit"],"dc:publisher":["Institutional Repository of Vilnius University"],"dc:relation":["https://epublications.vu.lt/object/elaba:308337699/308337699.pdf"],"dc:rights":["info:eu-repo/semantics/openAccess"],"dc:title":["Ranka rašyto lietuviško teksto atpažinimas /","Handwritten lithuanian text recognition."],"dc:type":["info:eu-repo/semantics/bachelorThesis"]},"updated_at":"2026-11-19T10:12:59Z"}