This repository is a comprehensive, curated "awesome list" of resources for Optical Character Recognition (OCR). It covers a vast spectrum of tasks, from basic deskewing and dewarping to advanced scene text recognition, document layout analysis, and handwritten text recognition. It serves as a valuable starting point for researchers and developers exploring OCR technologies, providing pointers to relevant papers, code, and datasets.

How It Works

The list is organized thematically into numerous categories detailing specific OCR sub-fields. These include document preprocessing (deskewing, dewarping), layout analysis, text detection and localization (scene text, arbitrary shapes), handwritten text recognition, table extraction, and a wide array of OCR engines. Each entry typically links to a GitHub repository, research paper, or dataset, facilitating structured exploration.

Highlighted Details

• Extensive coverage of scene text detection and recognition, featuring numerous papers and implementations for oriented and arbitrarily shaped text (e.g., CRAFT, EAST, TextBoxes++, PSENet, Mask TextSpotter).
• Dedicated sections for document analysis tasks like layout segmentation (e.g., LayoutParser, DeepLayout) and dewarping (e.g., DewarpNet).
• A broad catalog of OCR engines and frameworks, from established tools (Tesseract, EasyOCR) to deep learning approaches (MMOCR, PaddleOCR, CRNN).
• Inclusion of resources for handwritten text recognition (HTR), font detection, table extraction, and essential pre/post-processing steps.

Maintenance & Community

As a curated list, this repository's maintenance is handled by its contributors. The primary community and development efforts reside within the individual projects linked herein. Users should consult respective project pages for specific details on their development status, contribution guidelines, and community channels.

Licensing & Compatibility

This collection does not have a single overarching license; each linked resource is governed by its own license. Users must consult individual licenses for software, datasets, or models. Commercial use or closed-source linking compatibility will vary significantly across the diverse set of included resources.

Limitations & Caveats

This is a curated list, not a unified OCR system. The quality, maturity, and maintenance status of individual projects vary widely, and links may become outdated. Its breadth means it may not offer deep dives into every niche. Users must independently evaluate each resource for specific needs, including development status and potential adoption blockers.