Development of Braille Text Recognition System into Alphabet Using YOLOv8

Visually impaired individuals often demonstrate significant creative potential by using Braille as a medium of expression, allowing access without reliance on electronic devices. However, the majority of the general public faces difficulty in reading Braille due to limited familiarity with the script in daily life. Deep Learning represents a branch of artificial intelligence and machine learning that has made significant advancements. One prominent algorithm for real-time object detection is You Only Look Once (YOLO). Despite its efficiency, YOLO's application in image identification for Braille to alphabet conversion remains limited compared to other methods. This study aims to develop a Braille to alphabet recognition system using the YOLOv8 model. The system is expected to enhance conversion efficiency and facilitate information accessibility for visually impaired individuals through innovative technology. The evaluation results using precision, recall, and mean Average Precision (mAP) indicate high precision (0.942) and recall (0.816), demonstrating the model's accuracy in positive predictions while maintaining a good balance. The mAP50 is 0.892 and the mAP50-95 is 0.814, showing strong performance across various IoU thresholds.