DeFacing: need for automated wide comparison

The number of defacing and refacing* algorithms grows. Data acquisition changes. New analysis approaches etc.

Publications investigating effects targeting current defacing pipelines on some datasets already exist:
- https://cdn-akamai.6connex.com//645/1827//OHBM2020Poster_15919688426481638.pdf
- https://link.springer.com/content/pdf/10.1007%2Fs00330-019-06459-3.pdf

and similarly to alignment algorithms we don't know which is the best generally/in specific use cases. It would be nice to establish a fully automated pipeline which would grow with newer algorithms and possibly datasets

- could be without knowing ground truth -- just impact (difference) in the results of the pipeline (e.g. simple BET or some other morphometrics) across defacers
- could use some curated dataset(s) with ground truth known as e.g. pipeline + annotated datasets in https://mindboggle.info/ to judge the impact

Anyone interested to join @con/obc ?

* e.g. new refacing in AFNI, See poster 1030 on https://datalad-datasets.github.io/ohbm2020-posters/#/

Start by reviewing the linked publications and the OHBM 2020 poster, then compare the proposed ground-truth-free and curated-dataset approaches. Done means defining and establishing an automated pipeline that compares defacing/refacing algorithms across relevant datasets and measures their impact on downstream analysis.

Written by the indexing model from the issue text.