This document describes a study on currency recognition using machine learning. Specifically, it uses a Convolutional Neural Network (CNN) model to recognize Indian paper currencies based on their denomination. The researchers collected 600 currency images and cleaned the data. They then built a CNN classification model using features like color, size, texture, identification marks, etc. The model was implemented using TensorFlow Lite and achieved 95% accuracy on testing. The system is able to automatically detect currency denomination using image recognition and announce the amount using text-to-speech, providing a potential solution for currency recognition.