Start Here#
Three short, CPU-only lessons that take you from a fresh pip install
eegdash to a working PyTorch DataLoader over real BIDS-curated EEG
records. Each tutorial runs in under a few minutes, executes on the
docs CI matrix on every build, and has a matching YAML spec under
docs/tutorials/_spec/.
This category is the gateway to every other section in this gallery.
The core decoding workflow assumes you have already loaded a recording;
the features and evaluation tracks assume you understand windowing; the
transfer / foundation track assumes you can wire an EEGDashDataset
into a dataloader, with BIDS metadata handling per Pernet et
al. (2019).
What you will learn:
How to query the EEGDash index for datasets and records without downloading raw signals.
How to load one BIDS recording, inspect its sampling rate, channels, and events, and verify it before scaling up.
How to apply Braindecode-safe preprocessors and convert continuous signal into fixed-length windows.
How to wrap an
EEGDashDatasetin atorch.utils.data.DataLoaderand confirm one batch’s shape.The vocabulary – BIDS entities, record vs dataset documents, windows – you’ll see in every later tutorial.
Run the lessons in order:
plot_00_first_search.py– find datasets and records.plot_01_first_recording.py– load one recording and inspect it.plot_02_dataset_to_dataloader.py– build windows and a PyTorchDataLoader.
