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

about cvpr40 dataset #11

Closed HoshinoXxs closed 6 months ago

HoshinoXxs commented 6 months ago

thanks for excellent work when i run evluate.py in Triplet_LSTM/CVPR40, the model seems go wrong and unfitting

the model file is EEGStyleGAN-ADA-main/EEG2Feat/Triplet_LSTM/CVPR40/EXPERIMENT_29/bestckpt/eegfeat_all_0.9665178571428571.pth

the val files are: eeg_imagenet40_cvpr_2017/val

the result is as follow: [Epoch: 8042, Test KMeans score Proj: 0.0802407221664995] [0.02255639 0.04010025 0.01503759 0.02756892 0.03768844] cv_scores mean:0.02859032002115842

looking forward to a reply

prajwalsingh commented 6 months ago

Hi @HoshinoXxs , thank you for reading the work.

I cloned the code and re-run it on my system without any changes, and it gives 96% for the test and 96.7% for val (evaluate.py) of what was proposed in the paper.

I hope you are providing the correct dataset path (raw EEG, not 5_95 Hz). Kindly cross-check it. The only error I got was missing the directory; the rest of the results are accurate.

HoshinoXxs commented 6 months ago

Hi @HoshinoXxs , thank you for reading the work.

I cloned the code and re-run it on my system without any changes, and it gives 96% for the test and 96.7% for val (evaluate.py) of what was proposed in the paper.

I hope you are providing the correct dataset path (raw EEG, not 5_95 Hz). Kindly cross-check it. The only error I got was missing the directory; the rest of the results are accurate.

thanks for reply

after using the raw dataset, it works