Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.lumilexicon.com/llms.txt

Use this file to discover all available pages before exploring further.

Shelf Scan lets you photograph a whole shelf of books and identify them in bulk. It uses a multi-stage AI pipeline: spine detection, OCR, identification, and triage.

​
How it works

1

Photograph the shelf

Take a clear, straight-on photo of a bookshelf. Try to get the full shelf in frame.
2

Spine detection

Roboflow AI detects individual book spines in the image and draws bounding boxes around each one.
3

OCR

PaddleOCR (with AWS Rekognition fallback) reads the text on each spine.
4

AI identification

Claude identifies each book from the spine crop + OCR text. Title, author, publisher, and a confidence rating.
5

Triage

A separate AI agent classifies each book as high-value, moderate, or low-value to help you decide what’s worth researching.
6

Review & save

Review the results, correct any misidentifications, and save the ones you want to inventory.

​
When to use it

  • You’re at an estate sale and want to quickly assess a whole room of books
  • You’ve acquired a large lot and need to catalog it fast
  • You want a quick triage before deciding what to pull off the shelf
Shelf Scan costs 10 credits per photo because it runs multiple AI agents in sequence. Use Quick Snap for individual books — it’s much cheaper.