AgeSynth

Get Data

Description

AgeSynth is a fully synthetic database with aged samples as part of the release for the paper "Identity-Preserving Aging and De-Aging of Faces in the StyleGAN Latent Space" published at the IJCB 2025 conference. The provided aged latent vectors were generated with the official NVIDIA StyleGAN2 FFHQ 1024×1024 model. The dataset contains 20 000 fully synthetic facial identities with 10 additional variations each, edited along a continuous age axis so that younger or older versions can be synthesized. The data are split into two equal subsets created with different face-recognition backbones (IResNet-50 and EdgeFace-S). Every subset includes reference latents and embeddings, the aged and de-aged latents, and the parameters used to generate the references with the Synthetics-Disco package. These latents can be passed directly to the StyleGAN2 generator to obtain 1024×1024 images or adapted to downstream pipelines, and all supporting code and documentation are available on the Agesynth project page at https://www.idiap.ch/paper/agesynth/. The dataset is released for non-commercial research and education under a CC BY-NC-SA 4.0 licence.

Reference

If you're using this dataset, please cite the following publications

@InProceedings{luevano2024agesynth,
      title={Identity-Preserving Aging and De-Aging of Faces in the StyleGAN Latent Space},
      author={Luevano, Luis S. and Korshunov, Pavel and Marcel, S{\'e}bastien},
      booktitle = {International Joint Conference on Biometrics (IJCB 2025)},
      year = {2025},
      note = {Accepted for Publication in IJCB2025},
    }

@InProceedings{geissbuhler2024syntheticfacedatasetsgeneration,
      title={Synthetic Face Datasets Generation via Latent Space Exploration from Brownian Identity Diffusion}, 
      author={David Geissb\"uhler and Hatef Otroshi Shahreza and Sébastien Marcel},
      booktitle = {The Forty-second International Conference on Machine Learning (ICML)},
      year = 	 {2025},
      publisher =    {PMLR},
    }