Dataset Viewer
The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.
THINGS-EEG2, ATM preprocessed release (mirror)
This repository is a byte-identical mirror of a subset of
LidongYang/EEG_Image_decode
at revision dbe34bb2407164f70883c661c76664f0d596e522. It was made for the eeg2vis project so the data can be
fetched from one place. No file was changed. The upstream README.md is kept as ATM_README.md.
Contents
| Path | Content |
|---|---|
Preprocessed_data_250Hz/sub-01 … sub-10/preprocessed_eeg_training.npy |
Training EEG, 16,540 images × 4 repetitions, 63 channels, 250 Hz |
Preprocessed_data_250Hz/sub-01 … sub-10/preprocessed_eeg_test.npy |
Test EEG, 200 images × 80 repetitions |
ViT-H-14_features_train.pt, ViT-H-14_features_test.pt |
Image features from the upstream release |
images_set.tar.gz |
The THINGS stimulus images (training and test sets) with image_metadata.npy |
SHA256SUMS |
sha256 of every file above; check with sha256sum -c SHA256SUMS |
The EEG is already multivariate-noise-normalised (MVNN). The times array holds 300 values; the 250 stored samples
correspond to the last 250 of them, i.e. 0 to 0.996 s after stimulus onset.
Source and terms
- EEG: Gifford, Dwivedi, Roig and Cichy (2022), A large and rich EEG dataset for modeling human visual object recognition, NeuroImage 264, 119754. Original data: https://osf.io/3jk45/
- Preprocessing and features: the ATM release (Li et al., 2024, Visual Decoding and Reconstruction via EEG Embeddings with Guided Diffusion), licensed CC BY 4.0.
- Images: the THINGS database (Hebart et al., 2019, PLOS ONE). The images remain subject to the THINGS terms of use (https://things-initiative.org); use them for research only.
Please cite the original works when using this data.
- Downloads last month
- 27