SmartBookshelf methodology

SmartBookshelf processes a shelf photo in stages: detect visible spines, extract rough text, clean that noisy OCR, and return a reviewable book list instead of a raw text dump.

Pipeline overview

1. Locate visible books and likely spine regions.
2. Read title and author fragments from narrow vertical text.
3. Clean and match rough OCR into better candidate titles.
4. Return a structured list the user can review and save.

Why scans miss books

• Room-wide photos make spine text too small.
• Glare and shadow remove letters before cleanup begins.
• Stylized or low-contrast bindings are harder to match.