Skip to content

Latest commit

 

History

History
12 lines (8 loc) · 470 Bytes

README.md

File metadata and controls

12 lines (8 loc) · 470 Bytes

tessdata_best – Best (most accurate) trained models

This repository contains the best trained models for the Tesseract Open Source OCR Engine.

These models only work with the LSTM OCR engine of Tesseract 4.

See the Tesseract docs for additional information.

All data in the repository are licensed under the Apache-2.0 License, see file LICENSE.