snapshot images don't have unique images

[dschneiderch]

I discovered that snapshot images don't have unique filenames when creating #759. They are organized with subdirectories. This seems like a problem because the filename is used as the dict key in meta so you would lose all images except the last. Is it an impossible situation because of the way SnapshotInfo.csv is created?

This would be a problem for metadata_parser with images if include_all_subdirs=True and the user is expecting all images in subdirs.

maybe using the full path as the key would be better.

[nfahlgren]

In our real datasets, the image file names are unique because they each contain a unique image ID from the database. The sample data I included in the tests snapshot directory is the same snapshot replicated, but I should have changed the image names too.

That being said, your point is still a good one. I have seen datasets from other systems where the directories are unique but not the filenames.