Overview

FPT.AI Reader - Facematch

Introduction

FPT.AI Reader - Facematch provides state-of-the-art deep learning algorithms to compare the portrait in personal identification documents such as ID card, driver’s license, passport… with real face to verify whether they are the same person.

This is the official user guide that support technicians, programmers and end-users about FPT.AI Reader - Facematch.

This document includes:

• Introduction to the system: Features, Performance, and Accuracy
• Requirements for the input images of the system
• Usage and features of API

Application

FPT.AI Reader - Facematch API can be applied for automated data input process in various domains including finance and banking, telecommunications, travel, tourism and hospitality; in such use cases as bank account opening, credit card application processing, user registration, guest verification / reservation, off-site check-in, etc.

The System

Features

The main function of the system is to detect faces on photos and identify that is same person or not. Users can send 2 images to the system, and this system will detect faces on images and return is same person or not. In order to evaluate the trustworthy of the results, the similarity probability of the detected faces is also returned

Performance

The processing time of the system depends on the resolution of the images submitted by the user.

The average processing time is:

• Server without GPU: 20-25 seconds/ photo
• Server with GPU (NVIDIA T4): 0.8-1.2 sec / photo

Accuracy

The average accuracy of the system is more than 95.4% if the input image meets all requirements.