OCR (optical character recognition)
Tesseract is an open-source OCR engine (C++, Apache-2.0) that extracts text from images using LSTM-based and legacy engines, supporting over 100 languages and multiple output formats.
Same problem, same approach. Swapping one for another is a config change, not a rewrite.
Python-native pip install, torch-based OCR engine with 80+ languages; choose over Tesseract for easy Python integration without C++ compilation.
Ready-to-use OCR with 80+ supported languages and all popular writing scripts including Latin, Chinese, Arabic, Devanagari, Cyrillic and etc.
Python OCR/layout-analysis engine with reading order and table recognition; choose when you need layout-aware extraction beyond plain OCR.
OCR, layout analysis, reading order, table recognition in 90+ languages
active Python/PP-OCR toolkit with 100+ languages and PDF-to-structured-data; choose for broad language support plus OCR and document parsing.
Turn any PDF or image document into structured data for your AI. A powerful, lightweight OCR toolkit that bridges the gap between images/PDFs and LLMs. Supports 100+ languages.
Solves the same problem with a different architecture or at a different layer. Expect to rewrite the integration.
Swift-native OCR library for Apple platforms; choose when you need a simple Swift API instead of a C++ CLI/library.
Fast and simple OCR library written in Swift
An earlier or more limited way to do the job. Still the right call when you need something small, proven, or CPU-only.
narrower OCR engine specialized for receipt images; choose when your input is scanned receipts and you want a purpose-built, lightweight Python option.
An efficient OCR engine for receipt image processing.
research code for training a CNN-LSTM-CTC text recognizer; choose if you need a trainable TensorFlow model rather than a ready OCR engine.
CNN+LSTM+CTC based OCR implemented using tensorflow.
experimental Caffe-based OCR model implementing CNN+BLSTM+CTC; choose for Caffe workflow experimentation.
主流ocr算法研究实验性的项目，目前实现了CNN+BLSTM+CTC架构
TensorFlow CNN+LSTM CTC OCR model for training/custom text recognition; more limited than a full engine.
Tensorflow-based CNN+LSTM trained with CTC-loss for OCR
specialized math-formula OCR with LaTeX output; choose when you need formula recognition instead of general text OCR.
:art: 数学公式识别增强版：中英文手写印刷公式、支持初级符号推导（数据结构基于 LaTeX 抽象语法树）Math Formula OCR Pro, supports handwrite, Chinese-mixed formulas and simple symbol reasoning (based on LaTeX AST).
captcha-specific OCR model training pipeline; choose if your task is captcha character recognition and you can train your own model.
[验证码识别-训练] This project is based on CNN/ResNet/DenseNet+GRU/LSTM+CTC/CrossEntropy to realize verification code identification. This project is only for training the model.
These projects were analysed and named tesseract among their alternatives. The relationship is not symmetric — how tesseract rates them is a separate judgement, made when tesseract is analysed in its own right.
calls tesseract “choose when you need a lightweight, mature OCR engine in C++ with no GPU dependency.”
Contexts Optical Compression
calls tesseract “traditional pipeline OCR engine, choose for lightweight, offline, multi-language OCR”
Official code implementation of General OCR Theory: Towards OCR-2.0 via a Unified End-to-end Model
calls tesseract “full OCR engine; choose for embedding or fine-tuning your own models”
PandaOCR - 多功能OCR图文识别+翻译+朗读+弹窗+公式+表格+图床+搜图+二维码
calls tesseract “The C++ Tesseract engine itself; choose for maximum accuracy and language support, not limited to short codes.”
Fast and simple OCR library written in Swift
calls tesseract “Traditional OCR engine with LSTM, choose for a mature standalone tool with broad language support and no training effort.”
主流ocr算法研究实验性的项目，目前实现了CNN+BLSTM+CTC架构
calls tesseract “Classic lightweight OCR engine; choose for fast, offline extraction without deep learning dependencies.”
Unlimited OCR Works: Welcome the Era of One-shot Long-horizon Parsing.
calls tesseract “battle-tested legacy OCR engine, lighter weight but no deep-learning table/layout.”
OCR, layout analysis, reading order, table recognition in 90+ languages