A Robust Deep Learning Framework for Bangla License Plate Recognition Using YOLO and Vision-Language OCR
Researchers have developed a robust deep learning framework for recognizing Bangla license plates, integrating object detection with optical character recognition. The system utilizes a novel two-stage adaptive training strategy based on YOLOv8 for license plate localization, achieving 97.83% accuracy and 91.3% IoU. For text extraction, a Vision-Language OCR approach with a ViT + BanglaBERT model demonstrated a character error rate of 0.1323 and word error rate of 0.1068. This framework shows consistent performance across diverse environmental conditions, making it suitable for intelligent transportation applications. AI
IMPACT Enhances OCR capabilities for specialized scripts, potentially improving automated traffic management systems.