The Signature Verification System (SVS) is designed to enhance security and efficiency in signature-based authentication processes. By utilizing advanced machine learning models integrated into a mobile application, SVS provides an accurate and automated solution for verifying handwritten signatures. This system addresses the limitations of manual verification by offering a reliable and real-time alternative.
Objectives
Our project objectives are to automate signature verification and validate a user’s identity through signature comparison, enhancing security in applications like banking, exams, and access control. Training a deep learning model that compares two signatures and determines their similarity using Euclidean distance. Allow users to upload or capture their signature using a phone camera or file upload for instant verification. Creating a user-friendly frontend for registration, login, and verification.
Socio-Economic Benefit
Reduces manual labor and the time required to verify signatures, leading to faster processing in banks, offices, and exams. Lowers administrative costs by automating identity verification processes. AI-based verification minimizes human error and subjective judgment in signature matching. Ensures fair and unbiased verification, improving trust in organizational systems. Users can register and verify signatures remotely via smartphone or webcam, promoting accessibility in remote or rural areas.
Methodologies
Defined project scope, user needs, and functional requirements. Identified core features: registration, login, signature capture, verification, and result feedback. Chose a Siamese Neural Network (SNN) architecture for one-shot learning and image comparison. Collected and preprocessed signature datasets. Trained the SNN using TensorFlow in and evaluated model accuracy using positive and negative signature pairs. Developed User Interface with HTML, CSS, and JavaScript for user interaction. Created a Flask-based server to handle user authentication, signature uploads, and model predictions. Used SQLite to store user credentials and associated signature image paths securely.
Outcome
Built a complete web-based system that allows user registration, login, signature upload, and verification through a user-friendly interface.
Created automatic image variation features to enhance model accuracy and simulate natural signature differences.
Verified signatures with strong accuracy and reliability, successfully distinguishing between genuine and forged signatures.
Developed a lightweight yet secure user authentication system with a local SQLite database for storing user data and signature images.
