Deep Learning Based Face Recognition

Deep learning-based face recognition aims to automatically identify individuals from images, focusing on improving accuracy, robustness, and explainability. Current research emphasizes developing more robust models against challenges like aging, occlusions (e.g., masks), and adversarial attacks (e.g., morphing), often employing architectures like Siamese networks, autoencoders, and variations of ResNet and VGG. This field is crucial for security applications (e.g., border control, access control) and necessitates addressing ethical concerns such as bias and privacy through techniques like differential privacy and explainable AI methods that analyze frequency domain information and human-interpretable feature maps.