This builds on the latest neurovault fetcher code.
To run:
The first time you run, use python qc.py to fetch all images.
Any other time, use python qc.py --offline to avoid extremely slow re-fetching.
Things to note:
Deduping is done on metadata only.
A check on detected duplicates is done in raw voxels. ~90% are raw dupes. A few don't match on raw pixels, but differ simply by the sign. 1 or 2 images don't match on the raw data, but still look identical. Are are removed as dupes as well.
Console outputs are generated for these exceptional cases. I checked the first 5 manually to summarize the above.
This builds on the latest neurovault fetcher code.
To run:
python qc.pyto fetch all images.python qc.py --offlineto avoid extremely slow re-fetching.Things to note: