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

训练步数的问题 #37

Open
zhangdaxia1 opened this issue Nov 27, 2019 · 4 comments
Open

训练步数的问题 #37

zhangdaxia1 opened this issue Nov 27, 2019 · 4 comments

Comments

@zhangdaxia1
Copy link

请问作者用ICDAR13-15数据集训练(229张训练图像)训练步数设置为多少步为好?训练的也比较慢,用GPU可能一晚上只能跑几万步,请问是什么原因呢

@Pay20Y
Copy link
Owner

Pay20Y commented Nov 27, 2019

6万到10万步应该就差不多。速度慢的话可以先看一下gpu使用率和cpu使用率,然后看一下瓶颈在哪里。也有可能是我实现的不是很好的原因。。。

@zhangdaxia1
Copy link
Author

感谢作者回答!麻烦再问下想先训练识别部分是否数据的标注也要像端到端那样呢?还是只需要得到label不需要得到label的位置?

@Pay20Y
Copy link
Owner

Pay20Y commented Nov 29, 2019

最直接的方式是提供像端到端那样的标注,然后不训练检测分支。你也可以直接用crop好的图片直接来训练识别,但那样就得改一下数据读入那里,这样只不过是把label的位置设置为全图了。

@zhangdaxia1
Copy link
Author

哦哦,听懂了!感谢作者耐心的解答!!!

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

2 participants