Current objective is to apply the framework of King et al. 2024 to The Present movie on HBN releases on NEMAR.
- Optimize data format: loading eeglab .set without .fdt in mne load the entire data. Might have to save to another format. Or update mne-python function to be able to load info only from single .set file. Ask Arno
- Integrate braindecode to this repo
- Verify channel location extraction and 2D projection are correct
- Optimize mne.Raw loading with lazy loading
- Verify that all HBN NEMAR releases can be loaded
- Handle movie frames
- Upload movie frame images to the repo
- Explore image embedding models
- Optionally format the annotations in a DataFrame to be used in the future
- Set up training pipeline following the paper
- Discuss training methods for our dataset