Identification of Fake

Indian Currency using

Neural Network

The detection of fake Indian currency is a critical issue. Explore

how transfer learning can be used to tackle this problem
effectively.
Overview of Transfer Learning
Definition Benefits Techniques
Transfer learning is a It accelerates the learning There are various ways to
technique in machine process, improves perform transfer learning,
learning where knowledge generalization, and including fine-tuning pre-
from one problem is reduces the need for large trained models and using
applied to a different but amounts of labeled data. feature extraction.
related problem.
Transfer Learning for Image
Classification

1 Explanation

Transfer learning in image classification involves leveraging knowledge

learned from pre-trained models on large datasets.

2 Benefits

It allows for better feature extraction, faster convergence, and improved

accuracy in image classification tasks.

3 Case Studies

Explore real-world applications of transfer learning in image classification,

such as object recognition and facial recognition.
Detecting Fake Indian Currency
Challenges Transfer Learning Data Sources
Approach
Detecting fake Indian Incorporating large image
currency is difficult due to Adopting transfer learning datasets of genuine and
the increasing enables the use of pre- counterfeit currency aids
sophistication of trained models to identify in training effective
counterfeit techniques. patterns and anomalies in detection models.
counterfeit currency.
Experimental Results and Analysis
Dataset Description Performance Discussion of
Evaluation Metrics Results
Analyze the dataset
used for training and Evaluate the efficiency Interpret the
evaluation, including and accuracy of the experimental results,
the number of images, detection model using identify potential
label distribution, and metrics such as limitations, and propose
data preprocessing. precision, recall, and F1 strategies for
score. improvement.
OBJECTIVE OF THE PROJECT:

 The project's primary goal is to develop a deep learning model that predicts fake currency by
analyzing images of banknotes.

 By training the model on a large dataset, it learns to detect subtle differences between
genuine and counterfeit notes, focusing on features like texture, color, and design patterns.

 This technology enhances the accuracy and efficiency of counterfeit detection, helping to
protect the integrity of financial transactions and reduce economic losses caused by fake
currency.
EXISTING SYSTEM:

 The existing systems for coin detection and recognition are limited by smaller datasets that
often lack diversity in coin types, shapes, and environmental conditions.

 These limitations hinder the accuracy and robustness of models trained on such data,
particularly in assisting visually impaired individuals.

 The current datasets typically do not cover a wide range of coin denominations and
environmental variations, reducing their effectiveness in real-world applications.
DEEP LEARNING DOMAIN:

 Deep learning is a branch of machine learning that

uses artificial neural networks with many layers to
learn complex patterns from large datasets.

 These networks, inspired by the human brain,

automatically extract and understand features from
raw data, making them highly effective in tasks like
image recognition, natural language processing,
and speech recognition.
LITERATURE SURVEY:
Paper 1: CNN based Counterfeit Indian Currency Recognition Using Generative Adversarial
Network
Author: Shamika Desai , Atharva Rajadhyaksha , Anjali Shetty
Year: 2021
Description: Paper currency's value exceeds its intrinsic worth, making counterfeit currency
detection crucial. Human vision can't identify fake currency effectively, and counterfeiters use
increasingly sophisticated methods. Existing methods are complex and inaccessible to the
general public. This project uses Generative Adversarial Networks (GANs) with Convolutional
Neural Networks (CNNs) to identify Indian paper currency. GANs consist of a Generator (creates
fake currency images) and a Discriminator (identifies real vs. fake). This approach aims to
provide a user-friendly solution for detecting counterfeit Indian currency.
LITERATURE SURVEY:
Paper 2: INDIAN CURRENCY FAKE NOTE DETECTION SYSTEM USING DEEP NEURAL NETWORK
Author: VVenkata Raghu Rami Reddy, 2Dr N Syed Siraj Ahmed
Year: 2021
Description: The existing system uses Convolutional Neural Networks (CNNs) and image
processing to distinguish between genuine and counterfeit Indian rupee notes. This approach is
a response to the increasing use of color printing technology for counterfeiting. Unlike
traditional methods like Support Vector Machine (SVM), which may struggle with large datasets
and noisy data, CNNs excel in processing images and can accurately identify fake currency
notes. The system analyzes various attributes such as color, shape, paper width, and image
filtering to make a quick and automated determination of a note's authenticity. This method
doesn't require human intervention and provides clear outcomes, including the note's value
Paper 3: DETECTION OF FAKE CURRENCY USING IMAGE PROCESSING AND NEURAL NETWORKS
Author: S.Suneetha, T.SaiMeenakshi

year: 2023
Description: The existing system uses Convolutional Neural Networks (CNNs) and image
processing to distinguish between genuine and counterfeit Indian rupee notes. This approach is
a response to the increasing use of color printing technology for counterfeiting. Unlike
traditional methods like Support Vector Machine (SVM), which may struggle with large datasets
and noisy data, CNNs excel in processing images and can accurately identify fake currency
notes. The system analyzes various attributes such as color, shape, paper width, and image
filtering to make a quick and automated determination of a note's authenticity. This method
doesn't require human intervention and provides clear outcomes, including the note's value
and whether it's genuine or counterfeit, making it efficient and effective in real-world scenarios.
Paper 4: FAKE CURRENCY DETECTION USING CONVOLUTION NEURAL NETWORK
Author: Ayush Antre, Om Kalbhor, Pratik Jagdale
year: 2023
Description:
This research paper introduces a deep learning system using Convolutional Neural Networks
(CNNs) to accurately distinguish between real and counterfeit currency notes. The system can
process images in real-time, making it valuable for banks, financial institutions, and businesses. Its
high accuracy helps prevent the circulation of counterfeit currency, safeguarding the economy and
the public from financial losses.
Paper 5: The image dataset of Indian coins: A machine learning approach for Indian currency

Author: Yogesh Suryawanshi a , Vidula Meshramb , Kailas Patil a

year: 2024
Description:
The Comprehensive Image Dataset of Contemporary Indian Coins (CIDCIC) is a valuable resource
with 6,672 images covering 53 classes of Indian coins. It aids in improving coin detection and
recognition, particularly benefiting visually impaired individuals by helping them accurately
identify different coin denominations. CIDCIC addresses the limitations of previous datasets,
which often lacked diversity and size, making it a crucial tool in advancing technologies like
computer vision and AI.
 Sl.no Title Author Year Dataset Algorithm Advantage Disadvantage
used
1 INDIAN V. Venkata Raghu 2021 We utilised the SVM Aim to Used for only
CURRENCY FAKE Rami Reddy, 2Dr dataset from, (Support increase the small datasets.
NOTE DETECTION N Syed Siraj which contains Vector accuracy of
SYSTEM USING Ahmed 4002 pictures of Machine) determining Performance is
DEEP NEURAL Indian paper if a note is not well.
NETWORK money notes in genuine or
denominations of counterfeit Requires high
₹10, ₹20, ₹50, memory
₹100, ₹200, ₹500,
and ₹2000.
2 CNN based Shamika Desai, 2021 r Paper Currency GAN(Genera This approach offers
the potential for
Relatively high level of
technical expertise for
Counterfeit Atharva Detection System tive improved accuracy,
development and
accessibility, and user-
Indian Currency Rajadhyaksha Based on Adversarial friendliness compared maintenance,
Recognition Using Combined SURF Network) to existing complex
and inaccessible
potentially limiting its
Generative and LBP Features hardware-based widespread adoption,
methods, making it especially in areas with
Adversarial (2018). easier for common limited access to such
Network people to detect and
identify counterfeit
expertise and
currency. computational
resources.
 Sl.no Title Author Year Dataset Algorithm Advantage Disadvantage
used
3 DETECTION OF S.Suneetha, 2023 Kaggle open source Neural Large dataset Prediction
FAKE CURRENCY T.SaiMeenakshi, Networks can be accuracy will be
USING IMAGE V.SivaSaranMarut trained low
PROCESSING AND hi,
NEURAL
NETWORKS
4 FAKE CURRENCY Ayush Antre, Om 2023 dataset of real and GCONVOLU Large dataset can be
trained
Prediction accuracy will
be low
DETECTION Kalbhor counterfeit TION
USING currency notes and NEURAL
CONVOLUTION achieve high NETWORK
NEURAL accuracy in
NETWORK detecting fake
notes
5 The image Yogesh 2024 Comprehensive RESNET50 This dataset
overcomes the
which might introduce
variability that
dataset of Indian Suryawanshi a , Image Dataset of limitations of previous
complicates model
collections, offering a
coins: A machine Vidula Meshram Contemporary more comprehensive training.
learning Indian Coins and diverse resource
for advancing
approach for (CIDCIC) technologies in
computer vision and
Indian currency AI.
CONCLUSION:

 In conclusion, the research successfully predicts

fake currency using a deep learning model.
 However, the literature survey highlights the need
for improved algorithms to further enhance
accuracy.
 PROPOSED work will focus on developing more
robust models for even more reliable counterfeit
detection.