User guide#
The user guide explains how to get the most out of docTR on your own documents, one topic at a time. Each page is self-contained and full of examples you can copy.
The essentials#
Read PDFs, images, web pages and in-memory data into pages.
Choose the models and configure the end-to-end OCR predictor.
Navigate the Document, convert coordinates, visualize and save predictions.
Text, JSON, Markdown, AsciiDoc, HTML and hOCR, in reading order.
Beyond plain text#
Detect titles, figures and tables, and extract tables as row/column grids.
Read skewed scans, photos and pages in any orientation.
Detect several classes of fields (dates, amounts…) and read them in one pass.
Use the detection, recognition, layout, table and orientation models on their own.
Find logos, QR codes, barcodes and photos with the contrib module.
Run OCR on files from your terminal with doctr-cli.
When something goes wrong#
Solutions to the most common installation, input and accuracy issues.