allenai/olmOCR-7B-0725
allenai/olmOCR-7B-0725 is a 7 billion parameter multimodal model developed by AllenAI, fine-tuned from Qwen2.5-VL-7B-Instruct. It specializes in Optical Character Recognition (OCR) for document images, extracting text, equations, and tables while preserving document structure. The model is designed for accurate text extraction from PDF document pages, converting equations to LaTeX and tables to Markdown.
Loading preview...
olmOCR-7B-0725: Document OCR Model
olmOCR-7B-0725 is a 7 billion parameter multimodal model from AllenAI, fine-tuned from Qwen2.5-VL-7B-Instruct. Its primary function is advanced Optical Character Recognition (OCR) for document images, specifically designed to process single pages of PDF documents. The model leverages the olmOCR-mix-0225 dataset for its specialized training.
Key Capabilities
- Accurate Text Extraction: Reads and transcribes plain text from document images.
- Structured Content Conversion: Converts mathematical equations into LaTeX format and tables into Markdown format.
- Contextual Reading: Preserves sentence continuity across pages, handling partial sentences from previous or continuing to next pages.
- Handwriting Recognition: Capable of reading natural handwriting within documents.
- Metadata Integration: Utilizes additional document metadata, typically generated via the olmOCR toolkit, for enhanced processing.
- Image Input Specification: Expects document images with the longest dimension set to 1288 pixels.
Good For
- Document Digitization: Ideal for converting scanned PDF pages into structured, editable text.
- Academic Research: Useful for extracting content from research papers, including complex equations and data tables.
- Information Retrieval: Facilitates the creation of searchable databases from image-based documents.
- Automated Document Processing: Can be integrated into workflows requiring precise text and structure extraction from diverse document types.