Demo · desktop

ShelfSense

A digital library from a photo of your bookshelf.

ShelfSense turns a photo of a bookshelf into a searchable catalog. Claude reads the spines, Google Books fills in the details, and the books land on shelves in a 3D library you can browse.

I started it in December 2025 and couldn’t get the quality there, so it sat on the shelf for most of the year. In October 2026 I came back with one prompt and a stopping condition, and Claude spent about 55 minutes measuring, fixing and re-measuring until the results held up across multiple photos.

See the demo

The ShelfSense Study: a 3D library room with bookcases on both sides of a fireplace, under a chandelier. — The Study. Choose a shelf to step up to it.

The ShelfSense Sun Room: white bookcases around a window looking out over the water. — The Sun Room, with light that follows the time of day.

What’s interesting about it

Measured, not eyeballed

On my labeled test photos, it finds 99.4% of the books in close-up shots, with 98.1% precision. The first scorer reported 0% because it paired shelves by number, so fixing the scorer came before fixing the pipeline.

Never confidently wrong

A wrong book is worse than a missing one, so uncertain matches go to a review queue. Across three full evaluation runs, it auto-accepted zero wrong books.

A library you can walk into

The Study has a fireplace, a reading chair and a side table of books in progress. The Sun Room looks out over the water, with light that follows the time of day.

Questions

What is ShelfSense?

ShelfSense is a web app that catalogs physical books from a photo of a bookshelf. Claude reads the spines, Google Books adds the details, and the result is a searchable catalog shown as a 3D library.

How accurate is ShelfSense?

On close-up test photos it found 99.4% of the books with 98.1% precision, and it never auto-accepted a wrong match. The sample is small: four photos of one set of bookcases.