Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

ocr斜45角的名片 完全不识别?不支持纠偏?内部没做一个透视修正? #10233

Closed
gowy222 opened this issue Jun 23, 2023 · 5 comments
Assignees
Labels
expneeded need extra experiment to fix issue good first issue Good for newcomers status/close

Comments

@gowy222
Copy link

gowy222 commented Jun 23, 2023

直接浏览器首测 劝退? .. .直接上图
156

@shiyutang shiyutang added the expneeded need extra experiment to fix issue label Jun 29, 2023
@Gmgge
Copy link
Contributor

Gmgge commented Jul 4, 2023

根据我的测试简要的回答下:
首先是:paddle ocr 中对于倾斜角度的字符检测并未倾注更多的注意力,在数据增强中有关角度变换的增强有:设置了(-10,10)的随机倾斜,随机字符行贴图(可能是倾斜的)
如果有倾斜文本的需求,可以有一下意见参考:
1.预处理增加文本方向纠正模块;
2.字符检测模型进行训练,增加倾斜数据与更强旋转增强。
关于意见的可行性,我这边做了预处理之后,在应用的数据集中,基本纠正了旋转的问题。如果你有进一步优化实验想法,可以提问更细化的问题,这样方便有相关经验的人为你提供一些意见。

@shiyutang
Copy link
Collaborator

shiyutang commented Jul 4, 2023

可以给出预处理相关的代码 @Gmgge

我这边做了预处理之后

@Gmgge
Copy link
Contributor

Gmgge commented Jul 5, 2023

@shiyutang 因为涉及内部项目问题,暂时无法提供旋转纠代码。如果提问者有进一步的想法,可以提供更具体的思路,例如想要尝试文本检测的多角度增强训练,或者希望做旋转纠正预处理。

1.在针对文本检测的多角度增强训练方向,可以尝试修改训练的配置文件即可获得相对稳定的效果;
2.旋转纠正需要确定使用者所应用的图像场景,往往是传统图像处理方法,难以在通用的场景保持精度

如果需要通用场景的多角度文本检测能力,建议是重新训练模型

@shiyutang
Copy link
Collaborator

好的~

@shiyutang
Copy link
Collaborator

以上回答已经充分解答了问题,如果有新的问题欢迎随时提交issue,或者在此条issue下继续回复~
我们开启了飞桨套件的ISSUE攻关活动,欢迎感兴趣的开发者参加:PaddlePaddle/PaddleOCR#10223

@shiyutang shiyutang added the good first issue Good for newcomers label Jul 7, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
expneeded need extra experiment to fix issue good first issue Good for newcomers status/close
Projects
None yet
Development

No branches or pull requests

4 participants