Fake Currency other Country Currency streamlit in Python Projects

Fake Currency other Country Currency streamlit in Python Projects

Abstract
Detecting counterfeit banknotes from multiple countries is a critical requirement for international businesses, banks, and border control agencies. This project presents a Python-based Streamlit application for fake-currency detection that supports notes from other countries by combining robust image-processing techniques, optical character recognition (OCR), and machine learning. The system accepts camera or scanner images of currency, performs preprocessing to correct perspective and illumination, and analyzes security features such as watermarks, microprinting, holograms, serial-number formats, color/texture patterns, and security threads. A hybrid approach using handcrafted descriptors (LBP, HOG, keypoints) together with learned CNN features enables the model to capture both global and subtle, country-specific differences. The Streamlit frontend offers an intuitive interface for uploading images, selecting the currency/country, viewing a genuineness score, and highlighting suspect regions—making the solution accessible for non-technical users while supporting continuous model updates as new counterfeit techniques emerge.
Existing System
Current counterfeit-detection practices vary widely: high-end devices use multispectral sensors, magnetic and UV readers, or specialized hardware to check physical security elements, while many small businesses rely on manual visual inspection or inexpensive UV pens. Software-only methods often focus on a single currency and use simple template matching, color thresholding, or fixed-rule checks, which break down under rotation, scale, worn notes, or when handling multiple national designs. OCR-based serial number checks exist but are fragile across fonts and image quality. Moreover, many legacy software approaches do not accommodate the diverse security features present across different countries, nor do they provide an easy way to add new currency models—limiting portability, maintainability, and real-world robustness.

Proposed System
The proposed system implements a scalable, multi-country counterfeit detection pipeline in Python with a Streamlit user interface. Images are preprocessed with illumination normalization, perspective correction, and background removal; region-of-interest localization isolates key areas (portrait, watermark window, serial number, security thread). Feature extraction mixes handcrafted descriptors (LBP, HOG, ORB-like keypoints) with a compact CNN backbone trained per-currency or via multi-task learning to learn cross-currency representations. An OCR module validates serial-number formats and checks against simple format rules or optional online lookup. The classifier outputs a probability/genuineness score and overlays heatmaps to indicate suspect regions. The Streamlit app lets users pick the currency, run batch checks, view explanations, and submit flagged samples to a retraining queue—enabling continual learning. Additional measures include data augmentation to handle wear and rotation, threshold tuning per-currency to reduce false positives, and a light admin panel for adding new currencies or updating security-feature templates. Deployed locally or on a lightweight server, this solution balances accuracy, usability, and extensibility for multi-national counterfeit detection.

Customer Reviews (0)

Share Your Experience

Rating

Your Comment

Submit Review Cancel

No reviews yet. Be the first!