Bias Mitigation in Face Recognition Models

Face recognition technology has become very important today, from unlocking our phones to keeping places secure trough surveillance. Be- cause of this, it is crucial to have models that are both efficient and fair. We will focus on FaceNet [1], one of the top models for face recognition, and look at how it might be biased. We will analyze these biases and then try to improve the model by using techniques to mitigate them. Our goal is to make FaceNet more accurate and fair for everyone. By fine-tuning the model and implementing bias mitigation strategies, we aim to ensure that FaceNet performs well across different groups of people, providing a more reliable and inclusive face recognition system. This will help in making the technology more trustworthy and useful in various real-world applications.