Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deep Learning Based Face Recognition Method using Siamese Network

Enoch Solomon, Abraham Woubie, Eyael Solomon Emiru

TL;DR

The paper addresses the challenge of relying on labeled data for face verification by introducing an unsupervised Siamese framework. It generates training pairs without labels from two disjoint datasets using cosine similarity and trains a double-branch CNN encoder to perform end-to-end verification. Key contributions include a practical unsupervised pair selection procedure, a shared-weight Siamese architecture, and an end-to-end verification pipeline that achieves competitive performance on LFW with substantially less labeled data. The results demonstrate data efficiency, with the unsupervised method approaching—but not universally surpassing—fully supervised baselines, highlighting its potential for scalable face verification in data-constrained environments.

Abstract

Achieving state-of-the-art results in face verification systems typically hinges on the availability of labeled face training data, a resource that often proves challenging to acquire in substantial quantities. In this research endeavor, we proposed employing Siamese networks for face recognition, eliminating the need for labeled face images. We achieve this by strategically leveraging negative samples alongside nearest neighbor counterparts, thereby establishing positive and negative pairs through an unsupervised methodology. The architectural framework adopts a VGG encoder, trained as a double branch siamese network. Our primary aim is to circumvent the necessity for labeled face image data, thus proposing the generation of training pairs in an entirely unsupervised manner. Positive training data are selected within a dataset based on their highest cosine similarity scores with a designated anchor, while negative training data are culled in a parallel fashion, though drawn from an alternate dataset. During training, the proposed siamese network conducts binary classification via cross-entropy loss. Subsequently, during the testing phase, we directly extract face verification scores from the network's output layer. Experimental results reveal that the proposed unsupervised system delivers a performance on par with a similar but fully supervised baseline.

Deep Learning Based Face Recognition Method using Siamese Network

TL;DR

The paper addresses the challenge of relying on labeled data for face verification by introducing an unsupervised Siamese framework. It generates training pairs without labels from two disjoint datasets using cosine similarity and trains a double-branch CNN encoder to perform end-to-end verification. Key contributions include a practical unsupervised pair selection procedure, a shared-weight Siamese architecture, and an end-to-end verification pipeline that achieves competitive performance on LFW with substantially less labeled data. The results demonstrate data efficiency, with the unsupervised method approaching—but not universally surpassing—fully supervised baselines, highlighting its potential for scalable face verification in data-constrained environments.

Abstract

Achieving state-of-the-art results in face verification systems typically hinges on the availability of labeled face training data, a resource that often proves challenging to acquire in substantial quantities. In this research endeavor, we proposed employing Siamese networks for face recognition, eliminating the need for labeled face images. We achieve this by strategically leveraging negative samples alongside nearest neighbor counterparts, thereby establishing positive and negative pairs through an unsupervised methodology. The architectural framework adopts a VGG encoder, trained as a double branch siamese network. Our primary aim is to circumvent the necessity for labeled face image data, thus proposing the generation of training pairs in an entirely unsupervised manner. Positive training data are selected within a dataset based on their highest cosine similarity scores with a designated anchor, while negative training data are culled in a parallel fashion, though drawn from an alternate dataset. During training, the proposed siamese network conducts binary classification via cross-entropy loss. Subsequently, during the testing phase, we directly extract face verification scores from the network's output layer. Experimental results reveal that the proposed unsupervised system delivers a performance on par with a similar but fully supervised baseline.
Paper Structure (12 sections, 1 figure, 2 tables, 1 algorithm)

This paper contains 12 sections, 1 figure, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Proposed Method
  3. Training Pairs Selection
  4. Proposed Siamese Network
  5. End-to-End Face Verification System
  6. DNN Encoder
  7. Experimental Setup and Dataset
  8. Experimental Setup
  9. Datasets
  10. Results
  11. Results on LFW
  12. Conclusion

Figures (1)

  • Figure 1: The architecture of the proposed siamese network. FC denotes fully connected layer. In evaluation phase, decision scores for face recognition are obtained directly from the last layer of the network. This layer computes the final representation of the features extracted from the input face images, which is then used to make binary classification decisions regarding facial similarity.