Kawi Character Recognition on Copper Inscription Using YOLO Object Detection

Inscriptions are an important source for learning historical events that occurred in the past. Since the middle of the 8th century AD, in the era of the Indonesian kingdom, inscriptions were written using Kawi character. The problem of reading Kawi character on inscriptions occurs when the media used is experiencing interference. This research focuses on copper inscriptions and uses YOLO object detection to recognize Kawi character on copper inscriptions, especially those with patina. Dataset used for the training in this research were 1518 images (1311 train, 207 validation) that comes from 69 original images of Warungahan Inscription using data augmentation. This research using 657 classes in total. In the training, tested on the validation images, the model obtained a mAP of 97.17% (threshold =0.4). In the test, tested on test images, the model recognized Kawi character on the Warungahan Inscriptions properly, with an accuracy of97.93% (threshold =0.4).