Expand description
Cheap image-region classifier + OCR budget (native + WASM). Image-region classification for OCR routing (SRP / Strategy).
Decides whether a raster should go through table OCR, prose OCR, or be skipped (charts / photos). Keeps the gate in one place so native and in-memory recovery share the same policy.
Structs§
- Classifying
Region Router - Classifier that runs heuristics, then optionally consults a table model.
- Heuristic
Image Region Classifier - Cheap heuristic classifier (no ML). Wraps the existing raster heuristics in
crate::pdf::raster_table_ocrso policy lives behind one trait. - Null
Table Structure Model - No-op table-structure model (classical bordered-grid path only).
- OcrBudget
- Shared deadline tracker for OCR work across a document.
Enums§
- Image
Region Kind - What to do with a page image region.
Traits§
- Image
Region Classifier - Strategy for classifying a grayscale page region.
- Table
Structure Model - Optional ML / table-structure model plugged into the
Tablebranch.