8000 Question · Issue #1 · wlkjyh/dianxuan · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

Question #1

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
Open
triangle959 opened this issue Dec 8, 2023 · 3 comments
Open

Question #1

triangle959 opened this issue Dec 8, 2023 · 3 comments

Comments

@triangle959
Copy link
  1. train.py 保存的是权重文件，代码漏了保存模型，后面预测时加载的是模型文件
  2. 这个input_shape 5252 是怎么确定的，我看样本里有很多3232，如果输入两个不一样的shape可以进行训练不？
@wlkjyh
Copy link
Owner
wlkjyh commented Dec 10, 2023

很高兴能回答你的问题，第一个问题，模型其实就是权重，模型里包含了权重和层的信息，训练时候保存的就是模型，你可以看到save_weights_only=False选项，当为False时候就会包含层信息，第二个问题，训练前其实会对样本进行预处理，会将非5252的形状resize为5252，你可以看下预处理那个部分

@triangle959
Copy link
Author
triangle959 commented Dec 12, 2023 •
edited
Loading

我测试了结果有点不佳，感觉是预处理部分正样本:负样本=4:3导致匹配结果不佳，还有图像预处理部分这类验证码应该不需要加上噪声，或许加上几何变换进行数据增强会好一些
image

@wlkjyh
Copy link
Owner
wlkjyh commented Dec 12, 2023

我测试了结果有点不佳，感觉是预处理部分正样本:负样本=4:3导致匹配结果不佳，还有图像预处理部分这类验证码应该不需要加上噪声，或许加上几何变换进行数据增强会好一些
image

你可以尝试下，当时我测试时候效果还行

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
2 participants
@triangle959 @wlkjyh
0