Face ID – Training Experience

Face Detection & Recognition

LiteRT (TensorFlow Lite)

facenet

This application provides a guided training experience for building custom Face ID models based on FaceNet-style embedding architectures. The solution includes detailed instructions, recommended workflows, and tooling guidance for dataset preparation, face alignment, augmentation, and the generation of high-quality facial embeddings using contrastive or triplet-loss training—performed externally in the user’s preferred environment (cloud or local workstation). Once the model is trained, the application enables streamlined deployment and efficient inference on SECO edge hardware through TensorFlow Lite–optimized pipelines. It also provides tools for embedding validation, similarity scoring, threshold configuration, and the creation or update of local identity databases. This approach ensures maximum flexibility during training while guaranteeing fast, private, and stable on-device inference for Face ID applications.