Money is a legal tender in the buying and selling transactions of society. However, visually impaired people face visual limitations that affect their ability to recognize the value of Rupiah banknotes. Although Rupiah banknotes already have a blind code feature, this is often ineffective due to the condition of the banknotes. In this study, we use Artificial Intelligence technology, especially deep learning, to help visually impaired people recognize the value of Rupiah banknotes more easily. Our system is built using Convolutional Neural Network (CNN) technique with MobileNetV3 architecture and Single Shot Multibox Detector (SSD) algorithm. The results show that our system is able to operate well under various lighting conditions, including daytime and nighttime. Under sufficient lighting conditions, the system achieves an accuracy of between 80% to 95%. However, we acknowledge that in low-light or nighttime conditions, our system has problems detecting banknotes. Thus, this research contributes to the effort of improving the accessibility of visually impaired people in recognizing Rupiah banknotes, although we also recognize the need for more attention to address low lighting conditions in future development.