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.

Sign up for GitHub

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

Jump to bottom

你好，请问训练该模型要多长时间呢？在什么机器上呢? #13

Open
2018hello opened this issue Apr 3, 2019 · 4 comments
Open

你好，请问训练该模型要多长时间呢？在什么机器上呢? #13

2018hello opened this issue Apr 3, 2019 · 4 comments

Comments

@2018hello
Copy link

你好，请问训练该模型要多长时间呢？在什么机器上呢?
@yinchangchang
Copy link
Owner

titan xp， 一个晚上就可以了

@2018hello
Copy link
Author

2018hello commented Apr 4, 2019 •
edited
Loading

好的，谢谢，想问下，为什么分类的时候不用Linear需要用卷积层作为输出呢？
请问下如果直接用densenet的去做单字符的话，效果怎么样呢？

@yinchangchang
Copy link
Owner

Linear是全连接层，会从整张图像特征预测，Conv是从局部特征预测，就识别文字而言，局部特征效果更好，Linear反而会混淆不同文字的特征

DenseNet会差好几个百分点

@xiliu
Copy link

xiliu commented May 13, 2019 •
edited
Loading

titan xp， 一个晚上就可以了

你好，我用的是densenet官方预训练模型，训练数据是自己生成的，当字符集为100或1000时（class_num=100,或1000）时，收敛都都很好。当字符集扩大到6863时，训练7小时（2块1080ti）也未见收敛，batch_size =32 ,lr=0.001,loss一直在0.1-0.2之间。

请问可能会是什么原因？

期待你的回复，非常感谢

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
3 participants
@xiliu @yinchangchang @2018hello