Signature Recognition using Siamese Neural Networks

Abstract

The traditional role of a signature is to permanently affix a person’s uniquely identifiable, self-identification to a document as physical evidence of that person’s personal witness and certification of the content of the entire document, or a selected portion of it. Yet with the technology advancement and tools available now a day it is easy to catch the culprits with artificial intelligence and neural networks. The neural network we used is Siamese neural network which is an artificial neural network which takes in similar input vectors and produces comparable output vectors. We are trying to make a user friendly interface using modern web technologies and the most data science language python and a simple Flask backend using python, the model is trained as per the signature images uploaded to a form on a local server then the images are taken and fed to the Siamese neural network to predict if they are forged or not and the results are going to be displayed in a user friendly format in the homepage.