search

SPOClab-ca / dn3

The fusion of MNE and PyTorch for accelerated deep-neural-network based BCI-systems and Neurophysiology signal analysis.
BSD 3-Clause "New" or "Revised" License
64 stars 19 forks source link

Reading BENDR Datasets #83

Open dsb589 opened 2 years ago

dsb589 commented 2 years ago

Hello - I was wondering if anyone has seen this thread, and might be able to help? @kostasde Myself and others have had some challenges trying to get the BENDR model to read some of the datasets used for that project's various downstream tasks. Different error messages are raised depending on the extension of the file:

For the BCI_IV_2a dataset, downloaded from here, if the specified extension in downstream_datasets is -raw.fif, I get:AssertionError: datasets should not be an empty iterable If I change the specified extension in downstream_datasets to match the extensions of the files I have downloaded (-.gdf), I get:cannot reshape array of size 520975 into shape (57855,newaxis).

For the Kaggle ERN Dataset:, downloaded from here, if specified extension in downstream_datasets is -raw.fif, I get: AssertionError: datasets should not be an empty iterable

For the ERPBCI Dataset, downloaded from here, if the specified extension in downstream_datasets set to -.fif, I get:AssertionError: datasets should not be an empty iterable. If the specified extension in downstream_datasets set to -.edf, I get:ValueError: file does not start with a file id tag

If you have any suggestions, please let me know. Thanks!

jc-audet commented 2 years ago

Hey @dsb589, did you manage to fix some of those problems? I'm currently running into the same thing and would be very grateful for some guidance!

kostasde commented 1 year ago

Hello all,

I'm not sure if people are still working on these issues, I've put these repos on the backburner lately to focus on some other things. Hopefully I can come back to this at some point again soon.

The empty iterable issues sound like somehow, your configuration for the dataset is excluding all of the data. E.g. some parameter choice is making the configuration ignore every sessions/person. Otherwise, depending on where you get the data, the files are not always perfect. Sometimes, making sure that they can be opened in MNE (e.g. one session is actually a broken file) goes a long way for debugging.

dee

On Thu, Sep 22, 2022 at 4:58 PM Jean-Christophe Gagnon-Audet < @.***> wrote:

Hey @dsb589 https://github.com/dsb589, did you manage to fix some of those problems? I'm currently running into the same thing and would be very grateful for some guidance!

— Reply to this email directly, view it on GitHub https://github.com/SPOClab-ca/dn3/issues/83#issuecomment-1255542284, or unsubscribe https://github.com/notifications/unsubscribe-auth/AJUPK2GSGYGXMDC4WIHEH33V7TCAXANCNFSM54R5RE4A . You are receiving this because you were mentioned.Message ID: @.***>