OmnEEG (pronounce OmnI-I-G) allows you to feed seamlessly multiple large and heterogeneous EEG datasets into your PyTorch models.

[X] PyTorch dataset integration

[X] YAML config files (global + cohorts)

[X] HDF5 export

[X] Integrate as a Transform operator in the Dataset class (see this tutorial)

[X] Topomap generation (Bashivan et al. 2015; see this class)

[X] Extract spherical coordinates of sensors (see this class, these classes, and that library)

[ ] Spherical model (Yao 2001)

[ ] Surface template (Groß et al. 2001)

[ ] Volumic template (Gramfort et al. 2013)

[ ] Individual anatomy morphed onto a template (Avants et al. 2008)

[ ] Check Riemanian geometry approaches (Sabbagh et al. 2020)

[ ] Check T-PHATE method (code and paper) and beyond (e.g., GSTH)

[ ] Train a model for ploting different representations (e.g., a "cubic brain") of the data based on the latent space.

If you use this software, please cite:

@software{RamezanianPanahi_Dumas_OmnEEG_2025,
  author = {Ramezanian-Panahi, Mahta and Dumas, Guillaume},
  title = {OmnEEG: Simple EEG tokenizer with PyTorch datasets},
  year = {2025},
  publisher = {GitHub},
  version = {main},
  url = {https://github.com/brain2vec/OmnEEG},
  note = {last updated: 2025-08-23; accessed: 2025-10-27}
}