Implementation of "Longitudinal Latent Diffusion Models"

Generated sequences

Disclaimer:

• The code in lib is an adaptation from [1]. In particular, we added there the LLDM implementation in lib/src/models/lldm.
• The code in geometric_perspective_on_vaes is an adaptation from https://github.com/clementchadebec/geometric_perspective_on_vaes

[1] Chadebec, C., Vincent, L. J., and Allassonni `ere, S. Pythae: Unifying generative autoencoders in python–a bench- marking use case. Proceedings of the Neural Information Processing Systems Track on Datasets and Benchmarks, 2022

Setup

Create an environment, activate it and install dependencies:

python -m venv lldm_env
source lldm_env/bin/activate
pip install -r requirements.txt

All the notebooks in the folder notebooks experiments-xxx.ipynb are then directly runnable and replicate the figures/tables of the paper. If anything wrong, please feel free to reach out.