[PDF] Synthetic Face Datasets Generation via Latent Space Exploration from Brownian Identity Diffusion | Semantic Scholar (2024)

[PDF] Semantic Reader

This paper introduces the proposed methodology to generate a synthetic dataset, without the need for human intervention, by exploiting the latent structure of a StyleGAN2 model with multiple controlled factors of variation, and confirms that the generated synthetic identities are not data subjects from the GAN's training dataset.

The results show the feasibility of the proposed GANDiffFace, in particular the use of Diffusion models to enhance the (limited) intra-class variations provided by GANs towards the level of real-world datasets.

The reported evaluation results on five authentic face benchmarks demonstrated that the privacy-friendly synthetic dataset has a high potential to be used for training face recognition models, achieving, for example, a verification accuracy of 91.87% on LFW using multi-class classification and 99.13% using the combined learning strategy.

This work presents a three-player generative adversarial network (GAN) framework, namely IDnet, that enables the integration of identity information into the generation process and empirically proved that the IDnet synthetic images are of higher identity discrimination in comparison to the conventional two-player GAN, while maintaining a realistic intra-identity variation.

This work proposes a framework, ExFaceGAN, to disentangle identity information in pretrained GANs latent spaces, enabling the generation of multiple samples of any synthetic identity and demonstrates the generalizability and effectiveness of ExFaceGAN by integrating it into learned latent spaces of three SOTA GAN approaches.

IDiff-Face is proposed, a novel approach based on conditional latent diffusion models for synthetic identity generation with realistic identity variations for face recognition training that pushed the limits of state-of-the-art performances.

A Dual Condition Face Generator (DCFace) based on a diffusion model that enables DCFace to consistently produce face images of the same subject under different styles with precise control and provide higher verification accuracies compared to previous works.

It is shown that it is possible to synthesize data with minimal domain gap, so that models trained on synthetic data generalize to real in-the-wild datasets and train machine learning systems for face-related tasks such as landmark localization and face parsing.

An overview of the submitted face recognition models and reports achieved performance compared to baseline models trained on real and synthetic datasets is given and an outlook on the current state of the research in training face recognition models using synthetic data is presented.

This paper first explores the performance gap between recent state-of-the-art face recognition models trained with synthetic and real face images, and devise the SynFace with identity mixup (IM) and domain mix up (DM) to mitigate the above performance gap.

...

...