Implementation of related angular-margin-based classification loss functions for training (face) embedding models: SphereFace, CosFace, ArcFace and MagFace.

Face matching using deep learning (CNN embedding + triplet loss)

API returns face feature vector as a response. This API uses deep learning to generate face embedding 128 dimension vector using Keras on top of TensorFlow. Implementation is "FaceNet: A Unified Embedding for Face Recognition and Clustering". Inception-ResNet-v2 model.

Multi-IVE: Privacy Enhancement of Multiple Soft-Biometrics in Face Embeddings

FaceNetを用いた顔分類

This is the documentation regarding the personal project to calculate the value of the average face embedding vector across a large dataset of faces/photos.