Chandra
Chandra is a state-of-the-art OCR model that converts images and PDFs into structured HTML/Markdown/JSON while preserving layout information. It handles complex tables, forms, handwriting, and supports over 90 languages. The model is available under the Apache 2.0 license, and the model weights use a modified OpenRAIL-M license. Chandra can be used via a CLI tool, a Streamlit web app, or a vLLM server. It has been benchmarked against other OCR models and has shown excellent performance, especially in multilingual settings. The model has been trained on a diverse set of documents, including math, tables, scans, and multi-column layouts.
- Category
- AI/ML
- Pricing
- Open Source
Problem solved
Chandra solves the problem of extracting structured data from unstructured documents, such as images and PDFs, while preserving layout information. This is particularly useful for applications where data needs to be extracted from documents, such as document scanning, data entry, and document analysis. Chandra's ability to handle complex tables, forms, and handwriting makes it a powerful tool for a wide range of use cases.
Use cases
- Document scanning and data extraction
- Data entry and document analysis
- Automated form processing
- Handwriting recognition and transcription
- Multilingual document processing
Audience
- Developers
- Data Scientists
- ML Engineers
Pros
- Highly accurate OCR capabilities
- Supports over 90 languages
- Handles complex tables, forms, and handwriting
- Available under permissive licenses
Cons
- Requires significant computational resources for training and inference
- May require additional processing for certain types of documents
- Limited support for certain languages or scripts
Official website: https://github.com/datalab-to/chandra
Canonical LambdaBase page: https://www.lambdabase.com/tools/chandra