<![CDATA[The ability to recognize banknotes' nominal value is a fundamental skill in daily economic transactions. However, for low-vision individuals, this simple task poses a major challenge, risking transaction errors and fraud. This study aims to build a web-based application capable of detecting Rupiah currency nominals in real-time by integrating computer vision and Optical Character Recognition (OCR) as an independent accessibility feature. The method combines a custom object detection model based on the YOLO architecture via the Roboflow platform and Tesseract OCR for nominal text verification, which is then integrated with the Web Speech API for voice-based output (Text-to-Speech). The system test results indicate that the combined "Roboflow + OCR" approach significantly improves detection reliability compared to using the object model alone. The system achieved a classification accuracy rate of 94.5% under optimal lighting conditions, with an average Text-to-Speech response latency of 1.8 seconds. This implementation proves that the synergy of image processing and OCR can provide an effective and inclusive assistive technology solution for visually impaired groups in Indonesia.]]>
Copyrights © 2026
