Synthetic Face Datasets Generation via Latent Space Exploration from Brownian Identity Diffusion

David Geissbühler, Hatef Otroshi Shahreza, Sébastien Marcel
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:19112-19136, 2025.

Abstract

Face recognition models are trained on large-scale datasets, which have privacy and ethical concerns. Lately, the use of synthetic data to complement or replace genuine data for the training of face recognition models has been proposed. While promising results have been obtained, it still remains unclear if generative models can yield diverse enough data for such tasks. In this work, we introduce a new method, inspired by the physical motion of soft particles subjected to stochastic Brownian forces, allowing us to sample identities distributions in a latent space under various constraints. We introduce three complementary algorithms, called Langevin, Dispersion, and DisCo, aimed at generating large synthetic face datasets. With this in hands, we generate several face datasets and benchmark them by training face recognition models, showing that data generated with our method exceeds the performance of previously GAN-based datasets and achieves competitive performance with state-of-the-art diffusion-based synthetic datasets. While diffusion models are shown to memorize training data, we prevent leakage in our new synthetic datasets, paving the way for more responsible synthetic datasets. Project page: https://www.idiap.ch/paper/synthetics-disco

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-geissbuhler25a, title = {Synthetic Face Datasets Generation via Latent Space Exploration from Brownian Identity Diffusion}, author = {Geissb\"{u}hler, David and Otroshi Shahreza, Hatef and Marcel, S\'{e}bastien}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {19112--19136}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/geissbuhler25a/geissbuhler25a.pdf}, url = {https://proceedings.mlr.press/v267/geissbuhler25a.html}, abstract = {Face recognition models are trained on large-scale datasets, which have privacy and ethical concerns. Lately, the use of synthetic data to complement or replace genuine data for the training of face recognition models has been proposed. While promising results have been obtained, it still remains unclear if generative models can yield diverse enough data for such tasks. In this work, we introduce a new method, inspired by the physical motion of soft particles subjected to stochastic Brownian forces, allowing us to sample identities distributions in a latent space under various constraints. We introduce three complementary algorithms, called Langevin, Dispersion, and DisCo, aimed at generating large synthetic face datasets. With this in hands, we generate several face datasets and benchmark them by training face recognition models, showing that data generated with our method exceeds the performance of previously GAN-based datasets and achieves competitive performance with state-of-the-art diffusion-based synthetic datasets. While diffusion models are shown to memorize training data, we prevent leakage in our new synthetic datasets, paving the way for more responsible synthetic datasets. Project page: https://www.idiap.ch/paper/synthetics-disco} }
Endnote
%0 Conference Paper %T Synthetic Face Datasets Generation via Latent Space Exploration from Brownian Identity Diffusion %A David Geissbühler %A Hatef Otroshi Shahreza %A Sébastien Marcel %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-geissbuhler25a %I PMLR %P 19112--19136 %U https://proceedings.mlr.press/v267/geissbuhler25a.html %V 267 %X Face recognition models are trained on large-scale datasets, which have privacy and ethical concerns. Lately, the use of synthetic data to complement or replace genuine data for the training of face recognition models has been proposed. While promising results have been obtained, it still remains unclear if generative models can yield diverse enough data for such tasks. In this work, we introduce a new method, inspired by the physical motion of soft particles subjected to stochastic Brownian forces, allowing us to sample identities distributions in a latent space under various constraints. We introduce three complementary algorithms, called Langevin, Dispersion, and DisCo, aimed at generating large synthetic face datasets. With this in hands, we generate several face datasets and benchmark them by training face recognition models, showing that data generated with our method exceeds the performance of previously GAN-based datasets and achieves competitive performance with state-of-the-art diffusion-based synthetic datasets. While diffusion models are shown to memorize training data, we prevent leakage in our new synthetic datasets, paving the way for more responsible synthetic datasets. Project page: https://www.idiap.ch/paper/synthetics-disco
APA
Geissbühler, D., Otroshi Shahreza, H. & Marcel, S.. (2025). Synthetic Face Datasets Generation via Latent Space Exploration from Brownian Identity Diffusion. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:19112-19136 Available from https://proceedings.mlr.press/v267/geissbuhler25a.html.

Related Material