Baidu Releases Open VLM for Advanced Document OCR
The new PaddleOCR-VL model is built to parse not just text, but also the tables, formulas, and page layouts found in complex documents.
Baidu has released PaddleOCR-VL-1.5, a new open-source vision-language model specialized in advanced document analysis. Released under a permissive Apache 2.0 license, the model aims to move beyond simple text extraction to understand the full structure and content of complex documents.
Unlike traditional OCR tools that focus solely on converting characters, PaddleOCR-VL is designed to handle a wider range of document intelligence tasks. According to its release notes, its key capabilities include:
- Full-page layout parsing
- Table structure recognition and extraction
- Mathematical formula detection
- Chart and graph analysis
This makes the model particularly suited for applications in academic research, finance, and enterprise document management, where information is often presented in structured, non-prose formats.
The model is based on Baidu's ERNIE 4.5 foundation, extending its multimodal capabilities specifically for the document OCR domain. By open-sourcing this specialized tool, Baidu provides developers with a powerful component for building applications that can digitize and interpret intricate information from images and scans. The model is available now on Hugging Face for community use and development.
Sources
- Visit
PaddlePaddle/PaddleOCR-VL-1.5
Hugging Face
0 comments
No comments yet. Be the first to weigh in.
More in Vision-Language
Moonshot AI Releases Kimi, a Multimodal Coding Model
The new Mixture-of-Experts model from the Chinese AI company can generate code while also understanding visual inputs, a rare combination in open models.
Google Releases Open-Source DiffusionGemma 26B Model
The new 26B parameter model from DeepMind uses a diffusion-based architecture, a technique more common in image generation, to produce text.
MiniMax Releases M3, a Multimodal MoE Model
The new open-weight model from MiniMax AI combines vision, coding, and reasoning using a Mixture-of-Experts architecture.