Libyan Currency Recognition System to Assist the Blind Community with Machine Learning Techniques

While recognizing banknotes may seem simple to those with normal vision, it presents a significant challenge for individuals in the blind community, especially when dealing with paper currency. The importance of money in daily life necessitates real-time detection and recognition of banknotes for blind or visually impaired individuals to conduct their business transactions with confidence. In this paper, an object recognition system designed to assist visually impaired individuals in their daily business transactions is proposed. To aid the blind community in banknote detection and recognition, smart glasses equipped with an Espressif32 (ESP32) camera for currency detection have been implemented. The approach involves the utilization of Teachable Machine, a real-time classification algorithm trained on a custom dataset of Libyan banknotes. Once the algorithm identifies the banknote, the label is processed and converted into audio using Text-to-Speech (TTS) technology, providing the expected output. A dataset consisting of Libyan banknote images captured in various scenarios was assembled for the algorithm's training. Subsequently, the system's robustness was enhanced by applying various geometric changes to the images, allowing the construction of reliable training and validation sets. The effectiveness of the system was evaluated using a test dataset. Experimental results demonstrate the system's ability to rapidly and reliably detect and identify Libyan currency, achieving an impressive accuracy rate of 96%.