Find My + Computer Vision

Imagine if computer vision models were so lightweight that they could run continuously, inexpensively, and reliably on any device. This would unlock entirely new user experiences.

One compelling idea is a next-generation “Find My” system for locating lost items. Instead of relying solely on tags or specialized hardware, users could train their devices to recognize their own objects — anything from a backpack to a water bottle — using a small, efficient computer vision model.

The breakthrough is twofold:

1. Personalized object recognition: Users can easily teach their devices to identify their belongings with high reliability.

2. Private, cooperative search network: These object identifiers can be securely and privately shared across a pseudo-FindMy-style network. Other participating devices can then help locate your item — without exposing personal data or object details.

A prototype application was built to test this idea, combining on-device training, camera-based recognition, and a distributed privacy-preserving sharing mechanism. The early results show that this approach is not only feasible but could redefine how people keep track of their everyday items.