<![CDATA[The number of visually impaired individuals in Indonesia reaches 1.5%, or around 4 million people, who often face difficulties in recognizing the denominations of rupiah banknotes. Although Bank Indonesia has added distinguishing features to the banknotes, this method is less effective due to limitations in understanding or the physical condition of the currency. Object detection technology, such as YOLOv8, offers a solution thanks to its advantages in accuracy and speed. This research employs the CRISP-DM approach, which includes six stages: business understanding to understand the needs of visually impaired individuals, data understanding to study the characteristics of the 2022 rupiah banknote dataset, data preparation to prepare 5,435 images of 8 currency denominations (1,000, 2,000, 5,000, 10,000, 20,000, 50,000, 75,000, and 100,000), modeling by training the YOLOv8n model, evaluation to assess model performance using a confusion matrix, and deployment on an Android application capable of real-time currency denomination detection through the camera. The evaluation results show an accuracy of 0.98, a precision of 0.988, a recall of 0.993, an average mAP50 score of 0.994, and an mAP50-95 score of 0.955, indicating that this model is quite effective in helping visually impaired individuals recognize the denominations of rupiah banknotes.]]>
Copyrights © 2025
