Fake Currency Detection System Final Year Project with Source Code

User / Application Features in this Final Year Project

• Browser-based Streamlit interface
• Simple single-user application
• Upload Indian currency note image
• Supports JPG image upload
• Supports JPEG image upload
• PNG support can be enabled
• Preview uploaded currency image
• Validate uploaded file format
• Process image in memory
• Resize image to CNN input size
• Convert image to numerical array
• Normalize pixel values
• Expand image into batch format
• Load trained Keras model
• Run CNN prediction
• Classify image as Real Currency
• Classify image as Fake Currency
• Display prediction result
• Display success message for real currency
• Display warning/error message for fake currency
• Optional confidence score display
• Error messages for invalid inputs
• Works locally in a browser
• CPU-compatible prediction
• No separate frontend framework required
• No login required

Streamlit Interface Panels in this Final Year Project

1. Application Header Panel

• Displays project title
• Displays short project description
• Shows usage instruction
• Shows currency authentication disclaimer
• Mentions supported image formats
• Provides simple user guidance before upload

2. Image Upload Panel

• Opens system file picker
• Accepts image file upload

• Allows supported extensions such as:

  jpg, jpeg, png

• Temporarily stores uploaded image in memory
• Sends image to preprocessing module

3. Image Preview Panel

• Displays selected note image
• Helps user verify correct image
• Shows uploaded image before prediction
• Helps detect blurred, rotated, incomplete, or incorrect images

4. File Validation Module

• Checks file extension
• Checks MIME type
• Checks image decoding
• Checks corrupted image cases
• Checks unsupported file types
• Checks image dimensions where implemented
• Displays clear error message for invalid upload

5. Image Preprocessing Module

• Opens image using Pillow or OpenCV
• Converts image to RGB format
• Resizes image to model input size
• Converts image to NumPy array
• Converts image to float values
• Normalizes pixel values between 0 and 1
• Expands dimension for model input batch
• Prepares image for CNN prediction

6. Model Loading Module

• Loads trained model file:

  Fake-currency.keras

• Uses TensorFlow/Keras model loading
• Can cache loaded model using Streamlit
• Avoids repeated model loading on each interaction
• Displays error when model file is missing or incompatible

7. Prediction Engine

• Sends preprocessed image to CNN model
• Calls model prediction method
• Extracts output probability
• Applies binary classification threshold
• Maps model output to readable class label
• Ensures class mapping matches training class indices

8. Result Panel

• Displays final prediction result

• Shows:

  Prediction: Real Currency

  or:

  Prediction: Fake Currency

• Can display confidence percentage
• Can show AI-generated result disclaimer
• Can show model version in future enhancement

9. Error and Notification Panel

Handles:

• No image uploaded
• Unsupported file type
• Corrupted image
• Missing model file
• Incorrect image shape
• TensorFlow loading error
• Prediction failure
• Low-confidence prediction
• Unexpected server error

Deep Learning / CNN Features in this Final Year Project

• Image-based fake currency detection
• Binary classification

• Classes:

  Real Currency
  Fake Currency

• Uses Convolutional Neural Network
• Uses TensorFlow
• Uses Keras
• Learns visual patterns from currency images
• Extracts note texture features
• Extracts print-quality patterns
• Extracts colour and design patterns
• Extracts visual structures from note images
• Supports model training in Jupyter Notebook
• Supports training in Google Colab

• Saves trained model as:

  Fake-currency.keras

• Supports future transfer learning extension
• Supports future Grad-CAM explainability extension

CNN Model Workflow in this Final Year Project

• Input image resized to model-defined shape

• Example input shape:

  128 × 128 × 3

• Convolutional layers extract visual features
• ReLU activation adds non-linearity
• Pooling layers reduce feature-map size
• Flatten or global pooling prepares features for classification
• Dense layers combine learned features
• Sigmoid output supports binary classification
• Threshold commonly uses 0.5
• Output is mapped to Real or Fake class

1. Download or clone the project folder:

   git clone <your-repository-url>
   cd Fake-Currency-Checker

2. Verify required files:

   Deploy.py
   Fake-currency.keras
   requirements.txt

3. Create virtual environment on Windows:

   python -m venv venv
   venv\Scripts\activate

4. Create virtual environment on Linux/macOS:

   python3 -m venv venv
   source venv/bin/activate

5. Upgrade pip:

   python -m pip install --upgrade pip

6. Install dependencies:

   pip install -r requirements.txt

7. If requirements file is not available, install manually:

   pip install streamlit tensorflow opencv-python numpy pillow matplotlib scikit-learn

8. Ensure model file is in the same folder as Deploy.py:

   Fake-currency.keras

9. Run Streamlit app:

   streamlit run Deploy.py

10. Open application:

http://localhost:8501

11. To run on port 8051:

streamlit run Deploy.py --server.port 8051

12. Open:

http://localhost:8051

13. If streamlit command is not recognized:

python -m streamlit run Deploy.py

Most Important -

Click on here to download Dataset and place it into Indian Currency Dataset Folder
Click on here to download ML model and place it into root folder

This project has no login credentials because it does not include login, registration, admin panel, or authentication module.

Credential note:
The Streamlit interface can be accessed directly after starting the application