Currency Detection System for Visually Impaired Users

Overview

In an effort to improve accessibility for visually impaired individuals, this project focuses on creating a Currency Detection System that leverages machine learning models to identify and classify currency notes in real-time. The system is designed to provide audio feedback, ensuring that visually impaired users can interact with their environment and receive clear, real-time updates about the currency they encounter.

CHECK HOW MODEL WORKS

We are Still optimizing the model for Object Detection and Currency Identification for frames with no Currency , So please use an Image Consisting of any Indian Currency to test this Model

Introduction

Currency detection has long been a challenge for the visually impaired community. This project addresses that need by developing an intelligent application that detects and classifies currency using real-time camera feed. The system makes use of cutting-edge technologies like YOLO for object detection and Convolutional Neural Networks (CNNs) for classification. Once a currency note is detected, the system announces the denomination via audio to the user, facilitating greater independence and financial accuracy.

Problem Statement

With a large number of different currencies and denominations across the globe, it can be difficult for visually impaired individuals to determine the value of currency notes. A system that can recognize currency in real-time and speak the value to the user will significantly improve their daily experiences.

Proposed Methodology

This project uses two main models:

1. YOLO (You Only Look Once): Used to detect the presence of currency in the live video feed.
2. CNN (Convolutional Neural Network): Trained to classify the detected currency notes into specific denominations.

MODEL DEMO

We are working on developing an app in Flutter and will soon Update the app details in this page. Thank You !

The system works as follows:

1. The user activates the system via voice input.
2. The YOLO model detects whether a currency note is in the frame.
3. Once detected, the CNN model classifies the currency note.
4. The result is then announced to the user using text-to-speech technology.

Model Architecture

1. YOLO Model:

   • Used for detecting the presence of currency notes in the camera feed.
   • It was trained with a custom dataset that includes over 1,000 images of different currency notes from various angles, lighting, and backgrounds.
   • The model achieves an impressive 93.5% accuracy in detecting currency.

2. CNN Model:

   • Based on a DenseNet pretrained model and fine-tuned to predict the currency denomination (e.g., 10 INR, 20 INR, etc.).
   • Trained for 13 epochs and fine-tuned with an additional 5 epochs to achieve high classification accuracy.

Data Collection and Dataset

The dataset for this project includes images of various currency denominations and is sourced from publicly available datasets as well as custom image collection.

Technologies Used

• YOLOv5 for object detection.
• CNN based on the DenseNet architecture for classification.
• TensorFlow for deep learning model implementation.
• Azure for storing and processing the dataset.
• Text-to-Speech (TTS) for audio feedback.
• Flutter for mobile app development.

Results

• The models successfully detect and classify currency with high accuracy.
• Real-time performance allows for seamless interaction with users.
• Speech feedback enhances accessibility for visually impaired individuals.

Dataset Link

• Currency Dataset - A collection of images for training the detection and classification models.

Installation

Prerequisites:

Ensure that you have the following installed:

• Python 3.x
• pip (Python's package installer)

Install Required Libraries:

To install the required libraries, use the following command:

pip install -r requirements.txt

Setting Up the Project:

1. Clone the repository:

   git clone https://github.com/your-username/currency-detection

2. Navigate to the project folder:

   cd currency-detection

Usage

To use the currency detection system, you need to:

1. Set up the camera feed (ensure a webcam or smartphone camera is available).

2. Run the app.py script to start the application:

   python app.py

The app will activate the camera feed and start processing the video to detect currency. Upon detecting a currency note, the system will announce the denomination.

Screenshots

Built With

• OpenCV - For real-time computer vision processing.
• TensorFlow - For machine learning model development.
• YOLOv5 - For object detection.
• SpeechRecognition - For speech input functionality.
• gTTS (Google Text-to-Speech) - For speech output functionality.
• Flutter - For developing the mobile app.

Contributing

Feel free to fork the repository, submit issues, or open pull requests. Any contributions, bug reports, or feature requests are welcome!

Conclusion

This project demonstrates the potential of combining object detection and classification models to create an accessible solution for visually impaired users. The use of real-time detection and text-to-speech feedback provides a meaningful improvement in daily life, particularly when interacting with currency notes.

License

This project is licensed under the MIT License - see the LICENSE file for details.