search

MischaRauch / Stable_Diffusion_on_EEG_Signals

Using Stable Diffusion on the DEAP Dataset to argument new data and improve classification
17 stars 2 forks source link

Quantitative assessment of quality #1

Open nengwp opened 1 year ago

nengwp commented 1 year ago

hi, sir Interesting project. How should I quantitatively assess the quality of the generated EEEG signal?

MischaRauch commented 1 year ago

Hey, thanks for your interest in my project! Since this is part of a bachelor thesis I will be finishing this project at the end of next week, so by then you can expect to find a more in depth explanation of what I did and the obtained results in a form of a report here. Stay tuned for that.

To quickly answer your question: I am not working with the raw signal as input for the diffusion model. Therefore, I extract features from the signal and use these as input for the DM. The quality of the synthetic features and the real features are than verified by a closer look into the feature distribution of both real and synthetic data set using SDMetrics. And a k-fold approach ideally yielding similar results on both datasets. If you want to check out the usability of DM on generating real EEG signals I would recommend this paper from Szabolcs Torma, Dr. Luca Szegletes

nengwp commented 1 year ago

Thank you very much for your quick and accurate response and thank you for your suggestions.