GutenOCR:文档的 grounded 视觉-语言前端
GutenOCR: A Grounded Vision-Language Front-End for Documents
AI总结:
GutenOCR 通过微调 Qwen2.5-VL 模型,实现了文档中的 grounded OCR 和检测,提升了 OCR 得分和文本检测召回率,但存在页面线性化和公式布局方面的权衡。
摘要:
GutenOCR 是通过微调 Qwen2.5-VL-3B 和 Qwen2.5-VL-7B 获得的一组 grounded OCR 前端。所得到的单检查点视觉-语言模型通过统一的提示接口暴露阅读、检测和 grounding 功能。在商业文档、科学文章和合成 grounding 数据上进行训练,模型支持全页和局部阅读,具有行级和段级边界框,以及条件 ``x 在哪里?'' 查询。我们介绍了一种 grounded OCR 评估协议,并展示了 GutenOCR-7B 在 10.5K 预留的商业和科学页面上,其复合 grounded OCR 得分比其 Qwen2.5-VL-7B 基础模型提高了超过一倍(0.40 到 0.82)。在 Fox 和 OmniDocBench v1.5 上,我们的方法显著提高了区域级和行级 OCR 以及文本检测召回率,但也揭示了在页面级线性化、颜色引导 OCR 和公式密集型布局方面的权衡。
英文摘要:
GutenOCR is a family of grounded OCR front-ends obtained by fine-tuning Qwen2.5-VL-3B and Qwen2.5-VL-7B. The resulting single-checkpoint vision-language models expose reading, detection, and grounding through a unified, prompt-based interface. Trained on business documents, scientific articles, and synthetic grounding data, the models support full-page and localized reading with line- and paragraph-level bounding boxes and conditional ``where is x?'' queries. We introduce a grounded OCR evaluation protocol and show that GutenOCR-7B more than doubles the composite grounded OCR score of its Qwen2.5-VL-7B backbone on 10.5K held-out business and scientific pages (0.40 to 0.82). On Fox and OmniDocBench v1.5, our approach substantially improves region- and line-level OCR as well as text-detection recall, but reveals trade-offs in page-level linearization, color-guided OCR, and formula-heavy layouts.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。