diff --git a/docs/blog/posts/images/640-20241208103523542 b/docs/blog/posts/images/640-20241208103523542 new file mode 100644 index 00000000..442372f8 Binary files /dev/null and b/docs/blog/posts/images/640-20241208103523542 differ diff --git a/docs/blog/posts/images/640-20241208103536368 b/docs/blog/posts/images/640-20241208103536368 new file mode 100644 index 00000000..ca81305a Binary files /dev/null and b/docs/blog/posts/images/640-20241208103536368 differ diff --git a/docs/blog/posts/images/640-20241208103541160 b/docs/blog/posts/images/640-20241208103541160 new file mode 100644 index 00000000..f78003f4 Binary files /dev/null and b/docs/blog/posts/images/640-20241208103541160 differ diff --git a/docs/blog/posts/images/640-20241208103546381 b/docs/blog/posts/images/640-20241208103546381 new file mode 100644 index 00000000..82ba8eb9 Binary files /dev/null and b/docs/blog/posts/images/640-20241208103546381 differ diff --git a/docs/blog/posts/images/640-20241208103550187 b/docs/blog/posts/images/640-20241208103550187 new file mode 100644 index 00000000..064cb7e7 Binary files /dev/null and b/docs/blog/posts/images/640-20241208103550187 differ diff --git a/docs/blog/posts/images/640-20241208103554131 b/docs/blog/posts/images/640-20241208103554131 new file mode 100644 index 00000000..75c206f7 Binary files /dev/null and b/docs/blog/posts/images/640-20241208103554131 differ diff --git a/docs/blog/posts/images/image-20241208103232900.png b/docs/blog/posts/images/image-20241208103232900.png new file mode 100644 index 00000000..d377c6ce Binary files /dev/null and b/docs/blog/posts/images/image-20241208103232900.png differ diff --git a/docs/blog/posts/images/image-20241208103310417.png b/docs/blog/posts/images/image-20241208103310417.png new file mode 100644 index 00000000..c2fb1d08 Binary files /dev/null and b/docs/blog/posts/images/image-20241208103310417.png differ diff --git a/docs/blog/posts/images/image-20241208103329977.png b/docs/blog/posts/images/image-20241208103329977.png new file mode 100644 index 00000000..6423937e Binary files /dev/null and b/docs/blog/posts/images/image-20241208103329977.png differ diff --git a/docs/blog/posts/images/image-20241208103344517.png b/docs/blog/posts/images/image-20241208103344517.png new file mode 100644 index 00000000..ce7ec2ec Binary files /dev/null and b/docs/blog/posts/images/image-20241208103344517.png differ diff --git a/docs/blog/posts/images/image-20241208103400365.png b/docs/blog/posts/images/image-20241208103400365.png new file mode 100644 index 00000000..c8a40b93 Binary files /dev/null and b/docs/blog/posts/images/image-20241208103400365.png differ diff --git a/docs/blog/posts/ocr_evaluation.md b/docs/blog/posts/ocr_evaluation.md new file mode 100644 index 00000000..fa2cebad --- /dev/null +++ b/docs/blog/posts/ocr_evaluation.md @@ -0,0 +1,58 @@ +--- +title: 广州软件院开箱测评12款开源OCR,RapidOCR表现优异 +date: + created: 2024-12-08 +authors: [SWHL] +slug: guangzhou-software-institute-reviews-12-open-source-ocr-tools-rapidocr-stands-out +categories: + - General +comments: true +typora-copy-images-to:images +--- + +RapidOCR在12款开源OCR中表现优异 + + + +前几天,Joker告诉我看到一篇OCR测评的文章,并说咱家RapidOCR整体表现很不错。我特地去看了一下,发现是广州软件院发起的OCR测评,算是一个比较公允的测评方了。这篇文章是今年6月份写的,也不算太老,还是有一点参考意义的。 + +RapidOCR在12款开源OCR中也算得上可圈可点了。完整的测评报告,请移步 [AI测评 | 十二款开源OCR开箱测评,文字识别哪家强](https://mp.weixin.qq.com/s?__biz=MzA5MjU2NjYxNQ==&mid=2449494038&idx=1&sn=f144be7182839e22a421e0df9e9609f0&scene=21#wechat_redirect) 获取。因为测评的文章没有开启原创,不能转载,因此就在这里给出了链接,在此感谢广州软件院的测评工作,为大家选择合适的OCR模型提供了参考。 + +⚠️注意:下面的图文均是摘自[AI测评 | 十二款开源OCR开箱测评,文字识别哪家强](https://mp.weixin.qq.com/s?__biz=MzA5MjU2NjYxNQ==&mid=2449494038&idx=1&sn=f144be7182839e22a421e0df9e9609f0&scene=21#wechat_redirect) 中,如有侵权,请告知我删除。 + +我在这里只放一些关键的测评结论: + (1)印刷中文的综合测评结果为: + +![image-20241208103232900](./images/image-20241208103232900.png) + +(2)印刷英文的综合测评结果为: + +![image-20241208103310417](./images/image-20241208103310417.png) + +(3)变形字体的艺术字测评结果为: + +![image-20241208103329977](./images/image-20241208103329977.png) + +(4)自然场景的街景图片测评结果为: + +![image-20241208103344517](./images/image-20241208103344517.png) + +(5)手写中文的综合测评结果为: + +![image-20241208103400365](./images/image-20241208103400365.png) + +测评总结: + +![图片](./images/640-20241208103523542) + +**RapidOCR虽然基于PaddleOCR,但做了⼀定预处理和参数优化,因此在⾓度旋转180度和低对⽐度两个数据集中拉开了与PaddleOCR的差距;并且基于onnx推理格式的版本,在识别的速度上也较PaddleOCR有明显提升。**各个⼯具在中⽂印刷⽂字正常扫描、拍照图⽚中识别的准确度都很⾼。其中,扫描⽂档数据集,前三名的綜合评分都在99%以上,照⽚⽂档数据集中,前三名的综合评分都在98%以上。 + +![图片](./images/640-20241208103536368) + +![图片](./images/640-20241208103541160) + +![图片](./images/640-20241208103546381) + +![图片](./images/640-20241208103550187) + +![图片](./images/640-20241208103554131) diff --git a/overrides/main.html b/overrides/main.html index e1280da6..ad072b0c 100644 --- a/overrides/main.html +++ b/overrides/main.html @@ -1,5 +1,5 @@ {% extends "base.html" %} {% block announce %} -

🎉 RapidOCR获得2024中国互联网发展创新与投资大赛(开源)二等奖 → link

+

🎉 广州软件院开箱测评12款开源OCR,RapidOCR表现优异 → link

{% endblock %} \ No newline at end of file