Kinship identification using age transformation and Siamese network.

Area of Science:

• Computer Science
• Biometrics
• Artificial Intelligence

Background:

• Kinship verification from facial images is crucial for various applications.
• Traditional methods like convolutional neural networks and transfer learning struggle due to domain differences in training data.
• A novel approach is needed to accurately identify familial relationships using facial data.

Purpose of the Study:

• To propose a new technique for kinship identification using facial images.
• To address the limitations of transfer learning in kinship verification.
• To improve the accuracy of identifying familial relationships.

Main Methods:

• A Siamese neural network architecture was employed for kinship identification.
• An age transformation algorithm was developed to facilitate comparison of parent-child facial images across different ages.
• The proposed method compares transformed-age images rather than actual images.

Main Results:

• The proposed kinship identification technique achieved an overall accuracy of 76.38%.
• The results demonstrate the potential of combining Siamese networks with age transformation for kinship verification.
• The method shows promise compared to existing transfer-learning approaches.

Conclusions:

• The developed Siamese neural network and age transformation technique offers a viable solution for kinship identification.
• Further improvements to the Life Span Age Transformation (LAT) algorithm could enhance accuracy.
• This research contributes to advancing facial image analysis for biometric verification.