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
检测模糊文字的思路 #105
Comments
确实可以用你说的方案,不过建议还是做个简单的不确定性估计。这样比较简单。------------------ 原始邮件 ------------------
发件人: "terrancraft"<notifications@github.com>
发送时间: 2020年10月28日(星期三) 晚上6:38
收件人: "WenmuZhou/PytorchOCR"<PytorchOCR@noreply.github.com>;
抄送: "Subscribed"<subscribed@noreply.github.com>;
主题: [WenmuZhou/PytorchOCR] 检测模糊文字的思路 (#105)
|
您指的是用dropout计算识别文字模型的不确定度吗?具体方案不太清楚,望指教。 |
不确定性估计是一个技术名词。自己搜。------------------ 原始邮件 ------------------
发件人: "terrancraft"<notifications@github.com>
发送时间: 2020年10月30日(星期五) 下午5:46
收件人: "WenmuZhou/PytorchOCR"<PytorchOCR@noreply.github.com>;
抄送: "Tao Luo"<744351893@qq.com>;"Comment"<comment@noreply.github.com>;
主题: Re: [WenmuZhou/PytorchOCR] 检测模糊文字的思路 (#105)
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
大佬,看了你许多源代码,收益匪浅。
请教你一个问题,我有一个坑爹的task需要识别模糊字。
但学习模糊字的特征真的是太困难了,我有个相反的思路:
用大量高清的样本去训练模型,不让模型具备泛化能力,然后在推理阶段同样用高清图片识别文字内容(我的task可以具备这样的推理环境),理论上模糊字会识别成奇怪的内容,这样可以通过比较识别结果和原本文的差异来判断是否有模糊字。
我的task因为从头到尾预测阶段只有一个样本(只是为了判断一张发票上有没有印模糊或者印错的字,不需要考虑泛化性和鲁棒性), 您觉得这个方案可不可行。
The text was updated successfully, but these errors were encountered: