Photos app in iOS 10 can distinguish 7 facial expressions, generate 33 Moments categories, and detect 4,432 searchable objects
This is not documented anywhere, so I took the liberty to jot some of these down. Next time when testing, you would have a better idea...
Context & Ripple Effects
A week after Apple announced that Photos would cluster pictures by people and scene objects using on-device computer vision, this teardown puts hard numbers on what shipped: seven distinguishable facial expressions, 33 auto-generated Moments categories, and a 4,432-entry searchable object taxonomy — none of it documented by Apple itself.
The scale matters because these classifiers became load-bearing product surfaces rather than demos: two years later the same face and scene recognition powers the For You tab's sharing recommendations, and a decade on the lineage runs through Visual Intelligence's Siri mode inside the Camera app.
First-order effects
- iPhone users upgrading to iOS 10 get a fully searchable photo library with zero manual tagging — every one of the 4,432 recognized object types becomes a query term.
- Apple ships a facial-expression classifier into consumer hands without documentation, meaning developers and testers discover capability boundaries only by probing.
Second-order effects
- Google Photos faces pressure to match search granularity at the category level, since Apple's object taxonomy sets user expectations for what 'search your photos' means.
- Because the classification runs on-device, the feature works offline and keeps images local — turning privacy positioning into a competitive weapon against cloud-dependent rivals.
Third-order effects
- If the pattern holds, each generation of on-device vision becomes infrastructure for the next product layer — from organizing libraries (2016) to recommending shares (2018) to analyzing the live world through the camera (2025–2026), with the phone's sensor stack doubling as an AI input surface.
The trend: On-device computer vision is expanding from passive photo-library organization toward real-time camera-based intelligence, with Apple reusing each classifier generation as the substrate for new product surfaces.