Indian Currency Dataset

Published: 28-08-2020| Version 1 | DOI: 10.17632/48ympv8jjf.1
Venkataramana Veeramsetty,
Gaurav Singal,
Tapas Badal


New Indian currency denomination dataset is not available online so a new dataset is created using Moto X-Play mobile camera has a 21MP camera. The size of all the currency images in the dataset which are captured in landscape mode is 5344×3006. Whereas the size of all the currency images in the dataset which are captured in portrait mode is 3006×5344. A total of 4657 images are captured to create the dataset. All the currency notes which are acceptable in the market are used like old and new 10 rupee notes, old 20 rupee notes, old and new 50 rupee notes, old and new 100 rupee notes, new 200, 500, and 2000 notes. To improve the dataset size data augmentation is applied to the currency note images. The different types of augmentations used are Zoom, Rotate90, Rotate270, Tilt, Distortion, and Flip. Dataset after data augmentation contains a total of 11657 images.


Steps to reproduce

This dataset was developed as a sabbatical at Bennett University with support from SR University (Formerly SR Engineering College), Warangal under the project "Coinnet: platform-independent application to recognize Indian currency notes using deep learning techniques". This dataset was created using Moto X Play mobile having a 21MP camera. This dataset can be useful the researchers who are working on currency-related problems like currency classification and counterfeit currency detection.