Abstract
Transfer learning introduces the ability to perform deep learning models over a small set of data. This paper investigates the utilization of three fine-tuned Convolutional Neural Networks (CNNs), namely, Alexnet, Googlenet, and Vgg16. Alexnet and Googlenet consider as the state-of-the-art models in deep learning, while Vgg16 preference due to its depth. Each model was fine-tuned, trained, and tested over a dataset contains Bosnian Banknotes (BAM). The dataset covers 11 classes where 10 images were collected through mobile phone camera for each class. Alexnet showed a better performance in terms of completing the training while Vgg16 showed better performance in terms of accuracy as it achieved 100% compared to 95.24% for Alexnet. Googlenet showed less efficient performance by achieving 88.65%.
Highlights
At present, there are many techniques utilized for image classification, fingerprint-based recognition, iris, and ear recognition, etc
The method for conducting this study is mainly based on two phases; first one is dataset collection for the Bosnian Currency paper, while the second phase is dedicated for testing and training that dataset on Alexnet, googlenet and Vgg16 pre_trained Convolutional Neural Networks (CNNs) models
This paper provided an explanation of the utilization of the concept Transfer Learning for Banknote recognition
Summary
There are many techniques utilized for image classification, fingerprint-based recognition, iris, and ear recognition, etc. Recognition, such as that was proposed in [2] This approach based on Ensemble Neural Network (ENN) where each individual network in that ENN was trained using negative correlation learning. It anticipated for solving the problem of currency recognition by applying the proposed approach on dataset includes images of seven types of Bangladeshi currency TAKA. R-CNN was trained in multiple stages [5] Another approach was proposed for Myanmar paper currency recognition [6] where the proposed approach based on the textural feature of the paper currency. The rest of the paper organized as follows: Section 2 the recent research in this area, the proposed methodology is explained in Section 3, where the last two sections 4 and 5 demonstrate the achieved results and the conclusion respectively
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
