Have you considered using federated machine learning for your face detection system? Do you solely use pre-trained models, or have you thought about training own models in the future to achieve even better clustering of customer photos' faces?