ocr使用paddleocr问题 #116
Replies: 2 comments 5 replies
-
|
rapidOCR的参数和paddleOCR不同，增加了补边，修改了缩放规则也调整一些其他参数 |
Beta Was this translation helpful? Give feedback.
-
|
所以如果要使用表格识别算法，最好还是使用rapidOCR作为ocr工具嘛？这样对表格的提取效果也会提升 |
Beta Was this translation helpful? Give feedback.
-
|
emmm,是的，其实rapidOCR也支持单字识别，对于有可能检测框会跨表格单元格的情况，支持更好 |
Beta Was this translation helpful? Give feedback.
-
|
好的好的 谢谢大佬 |
Beta Was this translation helpful? Give feedback.
-
|
大佬还有个问题，目前我看表格识别算法里，用的有线无线那块，都会默认加载rapidocr_onnxruntime的ocr，即使我在外部使用rapid_paddle解析，进去还会初始化这个。最后导致GPU使用失败 请问这个咋解决呢？ |
Beta Was this translation helpful? Give feedback.
-
|
emmmm,好像是会自动引入，我理解都安装一下应该不会冲突吧？就是既安装了rapidocr_onnxruntime也安装了rapidocr_paddle |
Beta Was this translation helpful? Give feedback.
-
|
现在好了，刚刚又提醒 说GPU失效，最好删掉所有onnxruntime的包 。但是刚刚测试好了 |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
在表格识别算法中，使用了paddleocr输出的结果作为ocr_result，发现出来的teds变差了很多，请问这个是情况呢？一定要用rapidocr嘛？因为rapidocr的gpu版本底层应该也是paddleocr，想问问为什么会导致teds变差呢
Beta Was this translation helpful? Give feedback.
All reactions