// the find
DeepBCI/Deep-BCI
An open software package to develop BCI based brain and cognitive computing technology for recognizing user's intention using deep learning
A research code dump from Korea University's BCI lab covering deep learning approaches to EEG/NIRS signal classification — motor imagery, SSVEP, ERP, drowsiness detection, domain adaptation. Target audience is BCI researchers who want reference implementations of published papers, not engineers building production BCI systems.
Covers a genuinely wide range of BCI paradigms in one place — motor imagery, SSVEP, ERP, NIRS, ambulatory EEG — which is rare. The domain adversarial network and zero-training approaches are practically useful since cross-subject generalization is the real bottleneck in BCI. Some modules include pretrained checkpoints, which lets you verify the approach works before investing time in your own data. The inclusion of ear-EEG (ambulatory) alongside scalp EEG reflects actual deployment constraints that most BCI repos ignore.
This is academic code, not a library — no shared preprocessing pipeline, no common data format, each subfolder is an independent script pile with its own assumptions about input shape and file paths. Several folders contain placeholder files saying 'please request data', so you cannot run anything without emailing the authors. Last push was February 2024 and many modules still use TensorFlow 1.x checkpoint format (.ckpt.meta), which is a real pain to load in 2025. No requirements.txt or environment files at the repo level means you are on your own figuring out which TF/PyTorch version each subfolder expects.