Facial recognition technology plays a crucial role in numerous fields, from national security to social media platforms. Convolutional Neural Networks (CNNs) stand out as the most prevalent technique for facial recognition due to their impressive accuracy and adaptability in various scenarios. However, CNN-based learning faces considerable challenges, including high computational demands, a dependence on large volumes of labeled data, and significant storage requirements, which all impede its overall efficiency.
We aim to implement an innovative and efficient facial recognition approach utilizing structured neural networks grounded in Discrete Cosine Transforms (DCTs), achieving accuracy comparable to conventional CNNs. Our extensive and varied facial dataset guarantees robustness against diverse features, expressions, and lighting conditions. Furthermore, we will conduct a comprehensive comparison between our proposed neural network and traditional methods such as Eigenfaces, emphasizing their performance in facial recognition applications.