Швай, НадіяШпiр, Марiя2024-11-042024-11-042024https://ekmair.ukma.edu.ua/handle/123456789/32167This thesis addresses the challenges posed by data privacy laws, such as General Data Protection Regulation (GDPR), which limits the availability of datasets necessary for License Plate Recognition (LPR). It introduces a method to generate synthetic license plate images using diffusion models, specifically training a diffusion model on Ukrainian License Plates. These images were analyzed to demonstrate a high success rate in accurately replicating actual license plates. The findings reveal that using these synthetic, pseudo-labeled data as training datasets can increase LPR accuracy by 3% over traditional methods, showing the potential of diffusion models to produce realistic synthetic data for LPR applications, therefore avoiding the limitations set forth in the privacy regulations.ukdiffusion modelssynthetic datalicense plate recognitionGDPRprivacy regulationsбакалаврська роботаOther