search

eeyhsong / EEG-Conformer

EEG Transformer 2.0. i. Convolutional Transformer for EEG Decoding. ii. Novel visualization - Class Activation Topography.
GNU General Public License v3.0
389 stars 55 forks source link

Accuracy lower than that in the paper #35

Open wind-yong opened 3 months ago

wind-yong commented 3 months ago

I downloaded data BCI-IV-2a from the official website. Referring to your paper, I used the python MNE package for preprocessing. Running your Conformerd code on GitHub without filtering has an average accuracy of only 68.98%; If you add filtering, the accuracy is lower.May I ask why this is?

eeyhsong commented 3 months ago

There are many possible reasons such as pre-processing (segmentation, band-pass filter, z-score, etc), data augmentation, model parameters. I suggest to recheck the details.