AI BriefingTeleOCR: 90 points on real-world document recognition, 2nd in layout analysis
StarDoc-AI/TeleOCR
About the project
Converts documents from various environments, such as scans, photos, and screen captures, into digital data. It adopts a multimodal architecture based on Qwen2.5-VL to recognize not only text extraction but also tables and charts in a structured format.
It achieved a score of 90.72 on Real5-OmniDocBench, demonstrating its document recognition accuracy. In particular, it ranked among the top performers in the layout analysis category with a score of 80.14, showing strong capabilities in understanding complex document structures.
It is designed to maintain consistent performance across both digital documents and documents captured by camera. Licensed under Apache 2.0, it can be freely used in commercial services or research projects.
StarDoc-AI/TeleOCR
The original page has no description.
image-text-to-text
This introduction was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report errors, attribution issues, or removal requests via Contact.