<![CDATA[Banknotes and coins are valuable assets that are used as legal means of payment in everyday life. The value of these two types of money has been determined and is printed on each piece of banknote when used in transactions and trade. Even though currently banknotes can be recognized using technology such as ATM machines, these machines are only able to recognize the value of the largest currency owned by a country. Computers require digital images as input to display the information contained therein because computers do not have the ability of the human eye to directly recognize or calculate the objects they see. Therefore, techniques or methods are needed that aim to obtain information from digital images to facilitate human interpretation. This research aims to design a system for detecting banknotes in images using the Convolutional Neural Network (CNN) architecture, which is a form of deep learning. . The system also integrates image pre-processing using user-based manual annotation techniques in Python program code. Using the CNN method, a test was carried out to detect the nominal amount of money in the input image. Test results using 29 banknote dataset samples and 31 coin money dataset samples show that the two types of money are divided into two classes, namely paper and coins. From the training carried out on banknotes and coins, an average accuracy of 98% was obtained, showing good results. Repetition of the detection process also shows consistent output probabilities. However, there are several denominations of money that show high accuracy values, so it can be concluded that the labeling annotation method is thought to be less effective.]]>
