Top 10 Open-Source OCR Projects on GitHub Ranked by Stars

1. GOT‑OCR 2.0

GOT‑OCR 2.0 is an open-source end-to-end multimodal OCR model (1.43 GB) that can recognize text, mathematical formulas, molecular structures, charts, music scores, and geometric shapes. It has 7.2 K stars on GitHub.

https://github.com/Ucas-HaoranWei/GOT-OCR2.0

2. InternVL

InternVL, developed by OpenGVLab, is a large multimodal model comparable to GPT‑4V, capable of general image understanding and also compatible with OCR text extraction. It has 7.2 K stars.

https://github.com/OpenGVLab/InternVL

3. olmOCR

olmOCR, from AllenAI, focuses on linearizing PDF documents for large-language-model training, converting complex layouts into structured text. It requires a recent NVIDIA GPU with ≥20 GB RAM and ~30 GB disk space, and has 9.8 K stars.

https://github.com/allenai/olmocr

https://olmocr.allenai.org/

4. Zerox

Zerox (Omni‑AI) extracts text from PDFs, images, DOCX, etc., and outputs structured Markdown without any prior training. It leverages visual models such as GPT‑4o‑mini and has 10.3 K stars.

https://github.com/getomni-ai/zerox

https://getomni.ai/ocr-demo

5. Surya

Surya specializes in line‑level text detection, layout analysis, reading order, and table recognition, supporting 90+ languages. It excels at table structure extraction and complex document parsing, with 16.8 K stars.

https://github.com/VikParuchuri/surya

6. OCRmyPDF

OCRmyPDF adds a searchable text layer to image‑only PDFs using the Tesseract engine, supporting 100+ languages, batch processing, image de‑skewing, and cross‑platform deployment (Linux, Windows, macOS, Docker). It has 20.7 K stars.

https://github.com/ocrmypdf/OCRmyPDF

https://ocrmypdf.readthedocs.io/en/latest/

7. Marker

Marker converts PDFs, images, Office files, and EPUBs into Markdown, JSON, or HTML, preserving tables, formulas, and code blocks. GPU acceleration and optional LLM post‑processing improve accuracy; it has 22.8 K stars.

https://github.com/vikParuchuri/marker

8. EasyOCR

EasyOCR, from JaidedAI, is a PyTorch‑based library that returns extracted text, bounding boxes, and confidence scores for over 80 languages. It works on CPU/GPU and offers a simple API. It has 26 K stars.

https://github.com/JaidedAI/EasyOCR

9. Umi‑OCR

Umi‑OCR is a free, offline OCR tool for Windows 7+ x64 and Linux x64, requiring no internet connection. It supports batch processing, screenshot OCR, and various document types, with 30.8 K stars.

https://github.com/hiroi-sora/Umi-OCR

10. Tesseract

Tesseract is a widely used OCR engine originally developed by HP Labs and later open‑sourced by Google. It supports more than 100 languages and serves as the backend for many other projects. It has 65.3 K stars.

https://github.com/tesseract-ocr/tesseract

https://github.com/naptha/tesseract.js