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

Results #1

Open
teddyjkwang opened this issue Jan 23, 2018 · 10 comments
Open

Results #1

teddyjkwang opened this issue Jan 23, 2018 · 10 comments

Comments

@teddyjkwang
Copy link

Hi,thanks for the work.
Did you try on the dataset and how is the accuracy?

@LevinJ
Copy link

LevinJ commented Jan 24, 2018

Hi , @seragENTp , I am thinking of implementing that paper in tensoflow too. So am wondering if you have repeated the result in that parper? Thanks.

@lmolhw5252
Copy link

Hi, I use your code but it cannot convergence,and loss is always 2.29-2.3.
What's my problem? I don't change anymore.

@teddyjkwang
Copy link
Author

@lmolhw5252 I found the same problem.I thought the way to define the target of cross entropy is not right but not sure

@lmolhw5252
Copy link

@SwaggyK93 Can you fix it?

@LevinJ
Copy link

LevinJ commented Jan 30, 2018

Hi @SwaggyK93 , @lmolhw5252 , just thought you might be interested, I implemented the model in tensorflow and was able to reproduce the paper result.

Here is my code implementation.

@teddyjkwang
Copy link
Author

@LevinJ
THANKs so much for that!!

@teddyjkwang
Copy link
Author

哈哈谢谢老师 我会认真学习的

@lmolhw5252
Copy link

@LevinJ 非常感谢!您好我想请教您个问题,这篇论文中只是使用了3个数的MNIST数据集,并且设定切片个数为3,可是STN-OCR一文中,也是引用了这个方法,但是那篇论文好像没有切片个数的限制,请问您对这个有研究么?
“STN-OCR: A single Neural Network for Text Detection and Text Recognition”这个是那篇论文的题目。

@LevinJ
Copy link

LevinJ commented Jan 31, 2018

你好,我大致看了下论文,STN-OCR也是用了固定数目的切片。 文章有提到将来可能会研究由模型自动确定切片数目(或者说textline/characters 数目)。 你可以查看一下 "5.Conclusion"部分。

@lmolhw5252
Copy link

@LevinJ 非常感谢您!!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants