numediart/EEGLearn-Pytorch

some problems

dujiahei opened this issue · 1 comments

dujiahei commented

Thanks for your share!
I meet some problem as follows when I use it.

  1. The packages mkl-fft==1.0.15 in requirements.txt is not support Python3.7
  2. the shapes of self.Images are different, when different functons recall getitem（）.
    such as, (7, 202, 3, 32, 32) or (202, 3, 32, 32).
  3. the dimension of Images is not one, so the "image=Images[Patient_id==choosen_patient]" make some error.

Look forward to a new version code.

VDelv commented

Hi,

I am glad you find my repo useful.

  1. I am working on a conda environment with python 3.7.5 and mkl-ff==1.0.15 and didn't have any issues. I checked and pip install mkl fft==1.0.15 works.

  2. The shape of Images can be different, it depends on the model you consider, e.g. the mean of the seven images or choosing them separately (single frame or multi-frame approach in the article. Also explained in the notebook.

  3. In this case, it considers the first axis/dimension. If you prefer Images[id] correspond to Images[Id, :, :, :]. So it is not supposed to make any errors.

Hope it helps you,

If not, could you send the error printscreens?

Best,

VD