Facial recognition using PCA
The performance of facial recognition software has always suffered when it comes to unconstrained facial recognition. This is due to environmental variations such as changing illumination and occlusion, where part of the face is obstructed. Furthermore, faces in unconstrained facial recognition are...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
Nanyang Technological University
2020
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/140026 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | The performance of facial recognition software has always suffered when it comes to unconstrained facial recognition. This is due to environmental variations such as changing illumination and occlusion, where part of the face is obstructed. Furthermore, faces in unconstrained facial recognition are usually tilted at different angles and not frontal facing. It is a challenge to perform recognition accurately while taking these variations into account. In this project, I explore the application of Principal Component Analysis in combination with an Artificial Neural Network using a triplet loss function to learn low-dimensional feature representations of facial images, known as face embeddings. K-Nearest Neighbours classifier will then be used to perform verification and identification tasks. This method yields an improvement in performance over the traditional Eigenfaces approach where only Principal Component Analysis is applied and Euclidean distances between data samples and test samples are compared for classification. |
|---|