search

prajwalsingh / EEGStyleGAN-ADA

Pytorch code of paper "Learning Robust Deep Visual Representations from EEG Brain Recordings". [WACV 2024]
MIT License
27 stars 4 forks source link

EEGCLIP config file #7

Closed zihang-xu closed 6 months ago

zihang-xu commented 7 months ago

Hello~Thanks for your great work! I tried to reproduce the EEGCLIP with CVPR40 5-95hz and run the main.py, but could only get a KMeans score of about 0.04. It may be the problem with the config file. Could you please share this part? Or other guidance would be appreciated. : )

prajwalsingh commented 7 months ago

Hi @zihang-xu , thank you for reading our work.

Please refer to these issues for more information [ link, link ].

If you still have any doubts, do let us know.

zihang-xu commented 7 months ago

Here are some of my processes for reproducing the paper. I downloaded CVPR40 5-95hz zip and ran main.py. I tried with the paper, but the result was still not satisfactory. I think I missed some details. Could you point me in the right direction, please? Thanks again! 屏幕截图 2024-02-04 173359 屏幕截图 2024-02-04 175722

prajwalsingh commented 7 months ago

Hi @zihang-xu , here are the few things that I observed in the training code at this line [link], we are not passing the config file configuration so the network is using default initialized parameters that is:

in_channels=128, n_features=128, projection_dim=256, num_layers=1
. Make sure that you have the correct extracted data of 5-95Hz, and you can find the one that we have used [[link](https://iitgnacin-my.sharepoint.com/:u:/g/personal/19210048_iitgn_ac_in/ETX97Aw-USZd1HHU5mpIJmABPpl4fMCdEfkudJtuqRvNOw?e=3DFciF)], try running on this. Yo can also test the checkpoints that we have shared.
zihang-xu commented 7 months ago

Thank you very much! I will give them a try.

prajwalsingh commented 6 months ago

Hi @zihang-xu,

Did you run the code? Are there any updates regarding the results?

zihang-xu commented 6 months ago

Thank you very much for your concern and patience. I didn't follow up on this work continuously because of the holiday. I see the researcher is following up on the work about "EEGClip" at this link https://github.com/prajwalsingh/EEGStyleGAN-ADA/issues/9#issuecomment-1968347433. I encountered similar problems and https://github.com/prajwalsingh/EEGStyleGAN-ADA/blob/main/EEGClip/topkrecall.py doesn't print any metrics. I will keep an eye on the latest updates. Thanks again! 屏幕截图 2024-02-28 141126 屏幕截图 2024-02-28 143818

prajwalsingh commented 6 months ago

Hi @zihang-xu, thank you for your reply.

We are looking into it and will keep posting the latest findings in this [ thread ]. The problem arises due to messy code. We will clean it and update it on Git Hub.

prajwalsingh commented 6 months ago

Hi @zihang-xu , we have figured out the cause and explained it here.

zihang-xu commented 6 months ago

Thank you for your explanation and patience. I will read and study the code carefully.