Datasets:

---

imageomics

/

ToL-EDA

like 0

Follow

HDR Imageomics Institute 156

@thompsonmj can you add the licenses and update the file here?

@thompsonmj , as this has gotten rather bloated, here is a summary of where we're at on the mis-match between predicted-catalog.csv and the media manifest that should link to all these entries.

• eol_media_duplicates.csv: File with all images from predicted-catalog.csv that are duplicated in the media manifest, i.e., their Medium Source URLs match, but other identifiers are distinct.
• eol_cp_not_media.csv: File with all images from predicted-catalog.csv that do not match to media manifest.
• ToL_media_mismatch.ipynb: Notebook that explores and creates these files.

Note: Most mis-matched images and duplicates are in TreeOfLife-10M (i.e., they are in catalog.csv), but there are none in Rare Species dataset.

Three versions of media manifest:

• July 6 Media Manifest
• July 26 Media Manifest -- should be primary to match with our catalogs.
• Dec 6 Media Manifest

data/eol_files/media_content_not_catalog.csv is the CSV with all EOL content IDs that are in the July 26 Media Manifest, but don't match to EOL content and page IDs in predicted-catalog (matching via inner merge on EOL content & page IDs).

Idea is to pull all these images from their EOL Source URLs and match them against the images listed in eol_cp_not_media.csv, which has all the predicted-catalog images (tol ID, EOL content & page IDs) that don't match to media manifest.

Following commit 4886e4df, Licenses still missing Copyright Owners:

License Name
No known copyright restrictions    7389
cc-by-nc-sa-3.0                    4247
cc-by-nc                           3029
cc-publicdomain                     315
cc-by-3.0                            60
cc-0-1.0                             32
cc-by-sa-3.0                         28
cc-by-nc-2.0                          2

Note: only files with "No known copyright restrictions" for License Name are missing the URL.

Any unknown title has "not provided".

These issues are resolved in PR #10.