Memoria
Another way back to a photo.
Make a personal photo collection easier to browse, relate and organize through local metadata and image analysis.
How it works
Choose a library
The user selects a library source to index.
Build the index
Metadata and thumbnail services make large originals easier to browse.
Find connections
Image features support similarity, duplicate candidates and face-related grouping.
Review candidates
The user examines timelines, people, maps and candidate groups before acting.
The product idea
Find a photo through more than its date.
Use places, metadata and visual relationships as different ways into a personal photo collection.
- Discovery
Offer several paths back
Bring browsing, maps and grouping candidates together so different clues can lead to the same photograph.
- Precision
Separate exact matches from resemblance
Distinguish byte-identical files from visually similar candidates, keeping the strength of each signal clear.
- Traceability
Keep the analysis traceable
Store metadata and analysis locally, retaining feature sources when optional face models are used.
Independent projectLocal photo exploration; similarity suggests candidates and does not prove identity.
Stack
SwiftUI · AppKitSQLite3ImageIO · QuickLookApple VisionCryptoKit SHA-256Perceptual image hashesOptional ONNX Runtime · ArcFaceMapKit
Implementation details
Local photo exploration. Similarity is a candidate signal, not proof of identity.
- Photo library
- Read selected images and their available metadata.
- Library workspace
- Explore photos, maps and grouping candidates.
- Image analysis
- Detect faces and compute supported image features.
- Similarity checks
- Separate byte-identical files from visually similar candidates.
- Local index
- Keep extracted metadata and analysis results locally.
- Optional face model
- Use the optional model when present and retain backend provenance.