LOCAL VISION MODELS · PERSONAL PHOTO ARCHIVE
PhotoScanner.
Finding the words.
A tool for adding useful descriptions to a photo collection too large to describe one image at a time.
My archive contains around 80,000 photos. I developed PhotoScanner to turn local image understanding into a manageable process: scan, review and export, with the ability to stop and return later.
- Collection
- Around 80,000 photos
- Output
- Titles, descriptions & keywords
- Export
- XMP sidecars for digiKam
THE WORKFLOW
Suggestions first.
Review before export.
A local vision model looks at each photo and proposes a title, description and keywords.
The interface puts the photograph beside the suggested text. I can edit it, approve or reject it, or ask for a new description. This makes the output something I can work through, rather than an invisible batch of changes.
The descriptions are suggestions, not established facts. A model may miss details or infer something the image does not show, so review is an important part of the workflow.
WORKING AT SCALE
A queue that
can wait.
Large collections need more than a successful demonstration on a handful of images.
PhotoScanner scans folders and subfolders, saves its queue and results as it goes, and can resume later. It has already been used to process thousands of photos. Visible status and retry controls help separate successful results from files that still need attention.
The interface keeps completed files and remaining errors visible, so I can return to the items that need attention.
KEEPING THE ORIGINALS
Descriptions beside
the photographs.
Approved metadata is exported as XMP sidecar files for use in digiKam, without writing into the original images.
That separation lets me improve the archive’s descriptions while preserving the source photographs. The project combines model integration with the less visible work of queues, saved state, review and controlled export.
CONTINUE EXPLORING
Connected work.
Different applications.
Local AI Workbench
The wider work on models, tools and task reliability.
AI in practice
More projects connecting AI with practical needs.
IDEAS & PERSPECTIVES
Good ideas grow
through conversation.
I enjoy exchanging perspectives on people, technology and the things we learn by putting ideas into practice. If something here sparked your curiosity, let’s connect.
