DocFace+: ID Document to Selfie Matching Using Blockchain

Abstract

Varied activities in our manner needs us to verify who we are by showing our ID documents containing face pictures like passports and driver licenses, to human operators. However, this method is slow, labor intensive and unreliable. As such, AN automatic system for matching ID document photos to live face pictures (selfies) in real time and with high accuracy is needed. Throughout this paper, we have a tendency to propose DocFace+ to satisfy this objective. We have a tendency to 1st show that gradient-based improvement strategies converge slowly (due to the under fitting of classifier weights) once several categories have solely some samples, a characteristic of existing ID-selfie datasets. To beat this defect, to update the classifier weights, that permits quicker convergence and additional generalizable representations. Next, a combine of relation networks with part shared parameters square measure trained to hunt out a unified face illustration with domain-specific parameters. Cross-validation on AN ID selfie dataset shows that whereas a publically on the market general face intermediary.