# MOABB Roadmap

## Blocking bugs
- Handle FTP download for [Cho dataset](https://github.com/NeuroTechX/moabb/issues/77), with MNE PR
- [Shin2017](https://github.com/NeuroTechX/moabb/issues/80) with open source ack, push PR, ask Berlin if they could do something http://doc.ml.tu-berlin.de/hBCI/ (Morgan will ask if this is possible to update terms with the dataset authors) Alt. URL https://depositonce.tu-berlin.de//handle/11303/6271.2
- remove `moabb.run` from the README and provide minimal example

## Help wanted - direct contribution
- convert [Pedro's notebook](https://github.com/plcrodrigues/Workshop-MOABB-BCI-Graz-2019/tree/master/notebooks) (2.1, 2.2, 2.3 and 3.1) into python files and add it in [examples](https://github.com/NeuroTechX/moabb/tree/master/examples)
- Help triage bugs/PR and maintain the [Kanban](https://github.com/NeuroTechX/moabb/projects)
- Run evaluation using `moabb.run` and report results in [dedicated wiki page](https://github.com/NeuroTechX/moabb/wiki/Leaderboard-and-result-reports)

## Enhancements
- Leaderboard, paper with code like, in the doc or in the wiki or in a separate markdown page
- [X] switch to github actions instead of TravisCI
- add support for more ERP https://osf.io/thsqg/ and dataset, https://github.com/NeuroTechX/moabb/issues/87
- support different paradigm: cVEP, [auditory attention](https://zenodo.org/record/3377911), [neuromarketing](https://drive.google.com/file/d/0B2T1rQUvyyWcSGVVaHZBZzRtTms/view), affective BCI...
- mark tag project page, update with tasks for different profile (ML,
neuro) and level (beginner, intermediate, advance).
- add new dataset
- add new pipeline (like CCA)
- documentation
- contribution: could be test new, reproduce expe, make documentation
- remove link to https://docs.google.com/spreadsheets/d/1fQNFXGu1J1yJ9jFCer9EQQatjCPJWg7O-uCGF0Z4PiM/edit#gid=1596244060 and add a markdown page or something. Separate datasets and paper/bibliography, use https://github.com/meagmohit/EEG-Datasets
- build documentation, see ntx/eeg-notebook and PR from Erik Bjäreholt
- change moabb.neurotechx.com to neurotechx.github.io/moabb
- [X] adding a gitter chat: https://gitter.im/moabb_dev/community
- add CCA pipeline, see https://github.com/aaravindravi/Brain-computer-interfaces
- add TRCA SSVEP https://github.com/mnakanishi/TRCA-SSVEP
- remove notebooks
- https://github.com/NeuroTechX/moabb/blob/master/ROADMAP.md update
- standardize the name of datasets (get inspiration from open neuro and BIDS)
- new evaluation for small number of training samples
- new evaluation for cross-task, that is training one ERP/mental task for a subject and testing on another ERP/mental task for the same subject, e.g. training on left and right hand imagery and testing on feet and tongue.

## Big directions
Things that could be addressed during a code sprint

- transfer learning. Voir pour SSVEP https://iopscience.iop.org/article/10.1088/1741-2552/abcb6e/meta
- MNE support fNIRS, Shin2017 (berlin), add fNIRS support to MOABB, encourage submitting fNIRS datasets https://mne.tools/dev/auto_tutorials/preprocessing/plot_70_fnirs_processing.html see dataset http://doc.ml.tu-berlin.de/hBCI/ for fNIRS+EEG
- push BIDS EEG, ask data provider to comply to BIDS standard. See https://www.nature.com/articles/s41597-019-0104-8 and ongoing update on https://docs.google.com/document/d/1iaaLKgWjK5pcISD1MVxHKexB3PZWfE2aAC5HF_pCZWo/edit?pli=1# 2 official BIDS format and 2 non-official...
- link with Nemar https://sccn.ucsd.edu/projects/ https://sccn.ucsd.edu/projects/ Human Neuroelectromagnetic Data Archive And Tools Resource
- support other ML libs? TF/Pytorch ? for denoising using https://github.com/AllenInstitute/deepinterpolation
- evaluation of the preprocessing toolchain, like NARPS : https://openneuro.org/datasets/ds001734/versions/1.0.5
- add automagic-like (https://github.com/methlabUZH/automagic) preprocessing or with eye-blink detection : https://github.com/meagmohit/BLINK
- building documentation with moabb.github.io
- using https://www.datalad.org/ to keep track of datasets changes and reproductibility, any new pipelines could be tested against all existing datasets
- Connect with EEG notebooks, anyone could create a new local dataset that could be processed by MOABB pipelines
- setup pip package (pypi) or conda forge, best if done in CI
- new evaluation for mall