iOS SwiftUI MVP that helps users recover misplaced items using local camera breadcrumbs, reference photos, and on-device visual similarity search.
Built for HooHacks 2026, Breadcrumb answers:
Where did I last leave the thing I depend on?
The app keeps the promise narrow and honest:
- Observe a scene and let Breadcrumb auto-discover object candidates on device.
- Promote the candidates that matter into named memories when you want stronger identity continuity.
- Keep a local timeline of sightings, hand-linked interaction inferences, and last-known context.
- Add optional reference photos later if you want stronger re-identification for a specific object.
- Offline-first and local-first.
- Fresh installs work with zero registered objects and zero uploaded reference photos.
- Optional hosted refinement can be layered on later without breaking the offline core.
- No background surveillance recording.
Observe -> Saliency + Foreground Proposals + Vision Tracking + Hand Pose -> Candidate / Event Memory Graph
Promote Candidate -> Named Object Memory
Optional Reference Photos -> Stronger Re-identification Only When Needed
Breadcrumb.xcodeproj/
Breadcrumb/
App/
Features/
Register/
Capture/
Search/
Models/
Services/
Camera/
Search/
Storage/
Vision/
Shared/
Camera/
Views/
Resources/
docs/
dev/
research/
docs/dev/CODEX_HANDOFF.mdkeeps implementation notes and next-step validation details out of the main project surface.docs/research/contains the research memos, sources, and scorecards that shaped the MVP scope.
- Open
Breadcrumb.xcodeprojin Xcode on macOS. - Target is iOS 17+.
- The current implementation uses
AVFoundationfor live camera capture,Visionproposals and tracking for automatic discovery, and feature prints as a local re-identification aid. - This workspace was scaffolded in a non-macOS environment, so final device validation still needs to happen in Xcode on a Mac.