1 of 17

PRECISION FACE VERIFICATION WITH ALIGNMENT ENHANCEMENT

Shoaib Casseem

30/11/2025

2 of 17

AGENDA

Introduction

Face Recognition Pipeline

Primary Goal

App Background

Conclusion

Precision Face Verification with alignment enhancement

2

3 of 17

INTRODUCTION

  • Face recognition is a state-of-the-art employed throughout different areas of our lives.

​

  • The core principle of facial recognition for verification purposes is based upon the utilization of unique facial characteristics for secure authentication.

​

  • It presents a great scope in areas of security, user authentication and to carry out financial transactions.

​

​

​

Precision Face Verification with alignment enhancement

3

4 of 17

FACE RECOGNITION APPLICATIONS

Precision Face Verification with alignment enhancement

4

SECURITY

AUTHENTICATION

FINANCIAL TRANSACTION

SMART CITY APPLICATION

5 of 17

FACIAL RECOGNITION TECHNOLOGY

Precision Face Verification with alignment enhancement

5

Face Recognition

Face Identification

Face Verification

6 of 17

FACIAL RECOGNITION PIPELINE

Precision Face Verification with alignment enhancement

6

Face and Landmark Localization

Face Preprocessing

Feature Extraction

Face Recognition

7 of 17

FACIAL RECOGNITION PIPELINE

Precision Face Verification with alignment enhancement

7

Face and Landmark Localization

Face Preprocessing

Feature Extraction

Face Recognition

8 of 17

FACIAL RECOGNITION PIPELINE

Precision Face Verification with alignment enhancement

8

Face and Landmark Localization

Face Preprocessing

Feature Extraction

Face Recognition

Illumination Preprocessing

  • It is an efficient and effective approach in eliminating lighting variations.
  • Achieved using algorithm such as CLAHE

9 of 17

FACIAL RECOGNITION PIPELINE

Precision Face Verification with alignment enhancement

9

Face and Landmark Localization

Face Preprocessing

Feature Extraction

Face Recognition

Alignment Preprocessing

  • It is used to rectify face images into same canonical pose.
  • Usually achieved using basic transformations or deep learning approaches.

10 of 17

FACIAL RECOGNITION PIPELINE

Precision Face Verification with alignment enhancement

10

Face and Landmark Localization

Face Preprocessing

Feature Extraction

Face Recognition

Facial Detection and Cropping

  • The facial region is detected and cropped from background.
  • Achieved using algorithm such as MTCNN or SSD

11 of 17

PRIMARY GOAL

Improving face verification performance by performing facial alignment using a simple yet effective and innovative approach

Precision Face Verification with alignment enhancement

11

12 of 17

LIBRARIES USED TO CREATE THE APP

Precision Face Verification with alignment enhancement

12

OPENCV

NUMPY

MATPLOTLIB

KERAS

13 of 17

APP OVERVIEW

Precision Face Verification with alignment enhancement

13

    • USER IS PROMPTED TO UPLOAD ID
    • FACE REGION ON ID IS CROPPED AND ALIGNED
    • USER TAKES A SELFIE WITH WEBCAM
    • FACE REGION IS CROPPED AND ALIGNED
    • FACE VERIFICATION IS PERFORMED

14 of 17

HOW IS FACE ALIGNMENT PERFORMED?

Precision Face Verification with alignment enhancement

14

Right eye

Left eye

α

(x1,y1)

(x2,y2)

a

b

c

a = y1 – y2

b = x2 – x1

 

 

15 of 17

COMPARISON WITH OTHER SYSTEMS

Precision Face Verification with alignment enhancement

15

  • Up to 10% improvement in face verification performance as compared to system with no facial alignment.

​

  • The systems proves to be more successful in aligning and cropping face images as compared to Dlib’s 5 and 68-point facial landmark detector.

​

​

16 of 17

Precision Face Verification with alignment enhancement

16

CONCLUSION

  • Face alignment can be achieved using simple trigonometry and only 2 facial landmarks.

​

  • It a quick yet effective way to improve facial verification performance.

​

​

​

​

​

17 of 17

THANK YOU

Mohammad Shoaib Casseem

mohammad.casseem1@umail.uom.ac.mu

Precision Face Verification with alignment enhancement

17