HyperFace: Generating Synthetic Face Recognition Datasets by Exploring Face Embedding Hypersphere

TL;DR

HyperFace introduces a novel method for generating synthetic face recognition datasets by optimizing face embeddings on a hypersphere, enabling the creation of diverse and realistic training data that improve recognition performance.

Contribution

It formulates dataset generation as a packing problem on the face embedding hypersphere and solves it with gradient descent, advancing synthetic data creation for face recognition.

Findings

Models trained with HyperFace datasets achieve state-of-the-art accuracy.

Synthetic datasets generated by HyperFace improve face recognition performance.

The approach effectively balances inter-class and intra-class variations.

Abstract

Face recognition datasets are often collected by crawling Internet and without individuals' consents, raising ethical and privacy concerns. Generating synthetic datasets for training face recognition models has emerged as a promising alternative. However, the generation of synthetic datasets remains challenging as it entails adequate inter-class and intra-class variations. While advances in generative models have made it easier to increase intra-class variations in face datasets (such as pose, illumination, etc.), generating sufficient inter-class variation is still a difficult task. In this paper, we formulate the dataset generation as a packing problem on the embedding space (represented on a hypersphere) of a face recognition model and propose a new synthetic dataset generation approach, called HyperFace. We formalize our packing problem as an optimization problem and solve it with a gradient descent-based approach. Then, we use a conditional face generator model to synthesize face images from the optimized embeddings. We use our generated datasets to train face recognition models and evaluate the trained models on several benchmarking real datasets. Our experimental results show that models trained with HyperFace achieve state-of-the-art performance in training face recognition using synthetic datasets.