Currency Classification using DL in Raspberry Pi
₹9,999.00
Currency Classification using DL in Raspberry Pi
100 in stock
Description
Introduction
Real-time currency classification will be helpful for visually impaired people while doing shopping. Deep learning is well-known technology for visual application to do classification. In this project VGG16 Model is used with the fine-tuning method. Voice output will be given based on the recognized currency.
Abstract
It has Raspberry Pi as a core, which has USB/Mobile Camera and Speaker interfaced with it. A deep learning application opens the camera and process the frame to pass into the model to classify the currency. In this project, Indian currency is used, has 8 classes Ten, Twenty, Fifty, Hundred, Two Hundred, Five Hundred, Two Thousand & Background.
Existing system
In the existing system, classification is done through simple image processing to classify based on color, shape and other parameters.
Proposed System
In this proposed system, Deep learning is used with the help of Keras, contains several Models. Among those models, VGG16 Model is used with Transfer learning & Fine Tuning.
Connection Description
If the application needs USB Camera, it can be easily interfaced with the Raspberry Pi USB Port or it uses Mobile camera, Webcam android application should be installed on the Android mobile. Then through IP Configuration, streaming can be obtained from Mobile for an application.
Project Description
This project is capable of performing the real-time currency classification, includes processes like Training and Testing. In the process of training, VGG16 a pre-trained model will be loaded first, then the last layer will be replaced by our layer with new dataset of new class. Even a trained model is not trained for our classes, it can be able to classify our classes with the concept Transfer Learning. For further improvement of accuracy, Fine tuning is done by modifying the layers and retraining it.
Hardware Required
- Raspberry Pi
- USB Camera
- Power supply
- Speaker
Software Required
- Raspbian OS with libraries installed
- SD Card Formatter
- Etcher/Win32 Disk imager
- IPWebcam android app
Result
In the video demo, you can able to see the output obtained which classifies the currency with more than 90% accuracy with 8 classes.
Additional information
Weight | 1.000000 kg |
---|
Reviews
There are no reviews yet.