Tesseract — Open-Source OCR Engine

Overview

Tesseract is a command-line OCR engine and C/C++ library (libtesseract). Since version 4 it recognizes text line by line with an LSTM neural network, supports more than 100 languages through .traineddata files, and writes plain text, searchable PDF, hOCR, TSV, ALTO and PAGE XML. It reads images only (PNG, JPEG, TIFF and other formats Leptonica opens), not PDF, and it does not clean up bad scans much: most accuracy problems are solved before Tesseract runs, by fixing resolution, skew and contrast, and by choosing the right page segmentation mode.

Wrappers call the same engine: pytesseract shells out to the tesseract binary from Python, and tesseract.js runs a WebAssembly build in Node.js and the browser.

Instructions

Install the engine and language data

sudo apt install tesseract-ocr                       # Debian/Ubuntu: engine + eng + osd
sudo apt install tesseract-ocr-deu tesseract-ocr-fra # one package per language (3-letter code)
brew install tesseract                                # macOS: engine + eng + osd
brew install tesseract-lang                           # macOS: all other languages