01 / Overview
What it is and the problem it solves
Large photo collections make direct comparison slow, while deletion on mobile devices is particularly sensitive.
Windows desktop application. The public repository documents the product; a download will be added when a verified public release exists.
02 / How it works
How the system works
- 01
Local scanning and generation of a compact perceptual fingerprint per image to compare visual content rather than filenames alone.
- 02
Fingerprint and metadata persistence in SQLite to reuse previous work and process only new or modified files.
- 03
Match grouping and visual review before any destructive operation.
- 04
A native Rust bridge to interact with USB-connected Windows Portable Devices.
03 / My contribution
What I did
I designed the product, similarity pipeline, incremental persistence, interface and native Windows Portable Devices bridge.
Main decisions
- Use 64-bit perceptual hashing to drastically reduce the comparison space.
- Persist fingerprints in SQLite and recompute only new or modified files.
- Implement critical WPD operations in Rust and verify batches before deletion.
04 / Validation and outcome
How I checked the result
The flow separates detection, review and deletion so a similarity match never becomes an automatic removal. Batch operations are verified before execution and the incremental index avoids repeating the full analysis.
Project outcome
- Parallel and incremental analysis suitable for large libraries.
- Batch synchronization and deletion on USB phones with explicit checks.
- Installable Windows distribution.
05 / Lessons learned
What I learned during development
In a product that touches personal files, technical accuracy is not enough: the interface must explain why two images are considered similar and always keep the final decision with the user.