Inkspector: Signature Forgery Detection with Deep Learning

Inkspector is a deep learning project that detects forged handwritten signatures using a Siamese Convolutional Neural Network (Siamese CNN). By comparing a test signature against a known genuine sample, the model learns to distinguish subtle differences between authentic and forged signatures.

Project Overview

Forgery detection is a critical problem in banking, legal, and academic sectors. Traditional visual inspection methods are subjective and error-prone. Inkspector leverages deep learning to automate and improve the accuracy of signature verification.

Instead of treating signature classification as a binary task (real vs forged), Inkspector uses a pairwise comparison approach: it determines whether two signatures come from the same person.

Model Architecture

• Siamese Neural Network using shared CNN branches
• Each branch processes an input signature image and learns deep feature representations
• The model uses L1 distance to compare the features of a test signature and a known genuine sample
• Final prediction: same (genuine) or different (forged)

Dataset

The dataset used is Handwritten Signatures | Genuine and Forged from Kaggle.

• 150 genuine + 150 forged signatures (300 total)
• Format: Grayscale .png images

Dataset Structure:

• genuine/: 5 samples each from 30 individuals
• forged/: 5 forged samples for the same 30 individuals

File Naming Convention: NFI-XXXYYZZZ

• XXX: ID of the person who performed the signature
• YY: Sample number
• ZZZ: ID of the person the signature claims to represent

Examples:

• NFI-00602023: Person 006 forged a signature of person 023 → forged
• NFI-02103021: Person 021 signed their own signature → genuine

Performance

• Training Accuracy: 99.46%
• Validation Accuracy: 97.14%
• Custom confusion heatmaps and model visualization included

• The model in action below.

Project Structure

Inkspector/
├── sample_Signature/
│   ├── genuine/
│   └── forged/
├── dasv.ipynb           # Main notebook
├── Signature Forgery Detection.pptx
├── requirements.txt       
└── README.md

How to Run

1. Clone this repo

git clone https://github.com/jessechumo/inkspector.git
cd inkspector

2. Create and activate a virtual environment

python -m venv venv
source venv/bin/activate   # For Linux/macOS
venv\Scripts\activate      # For Windows

3. Install dependencies

pip install -r requirements.txt

4. Run the notebook

jupyter notebook dasv.ipynb

Features

• Siamese architecture for pairwise signature verification
• Heatmap visualizations to evaluate model behavior
• Plot of network architecture using Matplotlib
• Clean preprocessing with padding, normalization, and resizing
• Compatible across Linux (CPU) and Windows (GPU)

Future Work

• Expand dataset with real-world samples
• Add data augmentation for robustness
• Experiment with deeper or hybrid models
• Deploy as a web app for real-time signature verification

Credits

• Dataset: Divyansh Rai on Kaggle