Replies: 1 comment 6 replies
-
|
emmm,这个场景得先去掉印章，然后截取出表格区域，再做识别了，截取表格区域可以找版面检测的模型，去掉印章的话，好像没有特别好用的开源项目 |
Beta Was this translation helpful? Give feedback.
-
|
版面模型不在这个库里哈，这个库只负责表格结构识别，也就是说要求输入图片就是单个表格的图片，版面模型的话paddleX的docLayout应该最好，我们的RapidLayout或者RapidTableDetection应该能提取出来 |
Beta Was this translation helpful? Give feedback.
-
|
这边rapid系列的库都是很细分的，一般来说一个库只做一件事，ocr就是RapidOCR，表格的结构识别就是RapidTable,版面是RapidLayout |
Beta Was this translation helpful? Give feedback.
-
|
好的 我做一下RapidLayout的微调试一试 |
Beta Was this translation helpful? Give feedback.
-
|
想问一下RapidTableDetection和RapidLayout这两个库有啥区别呢，我如果只想提取表格部分 只用RapidTableDetection这个库是不是就可以了啊 |
Beta Was this translation helpful? Give feedback.
-
|
RapidLayout是用于版面分析的，会检测所有版面信息，RapidTableDetection只检测表格，且支持透视旋转的表格检测修正 |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
如这张表，有印章，就会把表格上边的一行识别为表格的第一行，以至于下边的单元格的行号多一，会识别第一行 “申请单位（公章），，”有印章的单元格内容也会识别错乱
Beta Was this translation helpful? Give feedback.
All reactions