百度发布新一代文字识别AI模型PP-OCRv5,仅0.07B,部分测试超GPT-4o

IT之家
Sep 13

IT 之家 9 月 13 日消息,百度于 9 月 10 日在 Hugging Face 发布新一代文字识别解决方案 PP-OCRv5

百度介绍称,PP-OCRv5 是一个为缓解大型视觉语言模型(VLMs)局限性而设计的专用 OCR 模型,它提供了一种高效、准确且轻量级的解决方案

PP-OCRv5 通过保持模块化、两阶段的流程,专门针对高速、精确的文本检测和识别,解决了大型 VLMs 的精确文本定位和边界框精度局限性问题。

PP-OCRv5 的亮点如下:

效率:该模型参数量仅为 0.07B,能够在 CPU 和边缘设备上实现更高性能,其移动版本在英特尔 Xeon Gold 6271C CPU 上每秒可处理超过 370 个字符。

性能:PP-OCRv5 在 OCR 特定基准测试中优于通用型 VLM 模型,如 Gemini 2.5 Pro、Qwen2.5-VL 和 GPT-4o,包括手写和印刷的中英文以及拼音文本。

定位:PP-OCRv5 旨在提供精确的文本行边界框坐标,这对于结构化数据提取和内容分析是关键要求。

多语言支持:该模型支持五种文字类型 —— 简体中文、繁体中文、英文、日文和拼音,并能识别超过 40 种语言。

PP-OCRv5 由四个核心组件构成:

图像预处理:处理图像的旋转和畸变,以标准化输入。

文本检测:识别图像中文本行的精确位置。

文本行方向:分类检测到的文本方向,以确保其正确对齐以进行识别。

文本识别:将每行文本中的字符解码为文本字符串。

Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.

Most Discussed

  1. 1
     
     
     
     
  2. 2
     
     
     
     
  3. 3
     
     
     
     
  4. 4
     
     
     
     
  5. 5
     
     
     
     
  6. 6
     
     
     
     
  7. 7
     
     
     
     
  8. 8
     
     
     
     
  9. 9
     
     
     
     
  10. 10