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

汉字识别训练的网络是哪个啊?lenet? #17

Open
tengshaofeng opened this issue Mar 10, 2017 · 5 comments
Open

汉字识别训练的网络是哪个啊?lenet? #17

tengshaofeng opened this issue Mar 10, 2017 · 5 comments

Comments

@tengshaofeng
Copy link

大哥您好,
非常赞赏你的工作。
能不能告知您训练的网络是基于哪个网络呢是以下这个吗?:https://github.com/JinpengLI/deep_ocr/blob/master/data/caffe_nets/lower_eng/lenet_train_test.prototxt

@yghstill
Copy link

首先要谢谢编主的分享,我也看了代码逻辑,也有这个疑问,直接使用lenet里面的lenet_train_test.prototxt和lenet.prototxt来进行训练测试吗?希望早日得到您的解答。谢谢

@JinpengLI
Copy link
Owner

不好意思,那么迟回复,工作太累不是很想回复。

不是这个lenet

https://github.com/JinpengLI/deep_ocr/blob/master/data/caffe_nets/lower_eng/lenet_train_test.prototxt

网格是基于这篇文章开发的

http://cs231n.stanford.edu/reports/zyh_project.pdf

中文训练的网格在百度云那个训练好的模型下载:

deep_ocr_workspace/data/chongdata_caffe_cn_sim_digits_64_64/lenet_train_test.prototxt

可以直接调用这个脚本来训练网络。

deep_ocr_workspace/data/chongdata_caffe_cn_sim_digits_64_64/train_lenet.sh

具体训练好的模型是在百度云可以下载,然后调用:

    base_dir = "/workspace/data/chongdata_caffe_cn_sim_digits_64_64"
    model_def = os.path.join(base_dir, "deploy_lenet_train_test.prototxt")
    model_weights = os.path.join(base_dir, "lenet_iter_50000.caffemodel")
    y_tag_json_path = os.path.join(base_dir, "y_tag.json")

@tengshaofeng
Copy link
Author

tengshaofeng commented Mar 14, 2017

非常详细谢谢。还有以下两个问题,能否得到您的指点:
1、我就是基于这篇paper里面的M7-1网络结构训练的,用tensorflow,为啥交叉损失一直在8下不去啊,一开始是8.3,循环n次还是8.2,而且调了各种优化器,学习率等。大神,你训练的时候会这样吗,你用了其他训练好的核参数对该网络初始化了吗

2、能否提供送入网络前的数据预处理代码呢?比如对比度最大化等

@yghstill
Copy link

感谢编主工作那么忙还回复,我去试试,我现在还在国内读研究生,也预见以后的工作会很忙碌,再次感谢。

@ilovin
Copy link

ilovin commented Oct 25, 2017

pdf下不了了,有没有下过的传一份?

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

4 participants