Reverse engineering Apple Photos - ML labels, quality scores and more

simonwillison.net

3 pointssimonw3 comments

I've been reverse-engineering Apple Photos and it turns out there is a WEALTH of exciting metadata about my photos stored in SQLite databases on my hard drive.

Not just machine-learning labeling of what's in the photos, but also calculated quality scores across a bunch of different criteria.

This is basically enough to get me to switch back to Apple's photos ecosystem for photo management.

Simon! This is great spelunking.

If you're frustrated by Photos corrupting its own library when it gets > couple hundred thousand assets, or you want to not be tied to Mac hardware, you might want to try PhotoStructure. (It has robust image deduping and tag inference, an open db schema, and CLI tooling).

Reply on news.ycombinator.com