CurrenSee is a deep learning-based currency recognition system that can identify the amount of any currency from images using a popular convolutional neural network called ResNet. It can detect seven types of currency: ten, twenty, fifty, one hundred, two hundred, five hundred, and one thousand. CurrenSee aims to assist visually impaired individuals. The dataset used for training CurrenSee includes data augmentation, such as changes in image orientation and light conditions, to increase its robustness. This system used ResNet-50, ResNet-101, and ResNet-152. These models achieved high accuracies, such as 96, 96.5, and 96%. Among these three models, ResNet-101 gives the highest accuracy. These results show the importance of the ResNet architecture for currency recognition and demonstrate practical application scenarios such as mobile apps, ATMs, and currency counters. In the future, this model will focus on the optimization of computational efficiency and an increase in the size of the dataset, which will include damaged or worn currency notes for improved robustness.