Post
19
How far has each EEG dataset on the Hub been checked?
BCI Report's open register has 487 dataset rows and 48 rows for models and papers, each with the furthest of five steps checked; for datasets: listed, read, traced, run, in the matrix. All 293 Hub rows are at step 1 (listed) so far: none of their files has been opened yet.
From a snapshot of the Hub on 19 October:
- 1,922 listings for 28 search terms and 20 tags; 737 are EEGDash mirrors, kept as one row.
- 368 uploads marked as brain-signal data after a classifier pass and a re-check. 146 of them carry no licence tag.
- 28 rows match an existing dataset by metadata. In 27 of them, the stated terms differ from the source's own.
- 15 match corpora that require a signed agreement, a request or a registration; 10 of those can be downloaded from the Hub without a request.
Matches use metadata only: names, sizes, counts. None of the matched uploads' files is downloaded, an uploader may hold permissions the register cannot see, and every row has a correction route, open to uploaders and dataset authors.
Write-up: https://huggingface.co/blog/Twu31/what-we-found-listing-the-eeg-datasets-on-hugging
Register: https://bci.report/datasets/register/
Data: https://bci.report/data/open-register-20261028.json
BCI Report's open register has 487 dataset rows and 48 rows for models and papers, each with the furthest of five steps checked; for datasets: listed, read, traced, run, in the matrix. All 293 Hub rows are at step 1 (listed) so far: none of their files has been opened yet.
From a snapshot of the Hub on 19 October:
- 1,922 listings for 28 search terms and 20 tags; 737 are EEGDash mirrors, kept as one row.
- 368 uploads marked as brain-signal data after a classifier pass and a re-check. 146 of them carry no licence tag.
- 28 rows match an existing dataset by metadata. In 27 of them, the stated terms differ from the source's own.
- 15 match corpora that require a signed agreement, a request or a registration; 10 of those can be downloaded from the Hub without a request.
Matches use metadata only: names, sizes, counts. None of the matched uploads' files is downloaded, an uploader may hold permissions the register cannot see, and every row has a correction route, open to uploaders and dataset authors.
Write-up: https://huggingface.co/blog/Twu31/what-we-found-listing-the-eeg-datasets-on-hugging
Register: https://bci.report/datasets/register/
Data: https://bci.report/data/open-register-20261028.json