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

长文本的效果忒儿差劲?这个怎么解决? #47

Closed
liuzhuang1024 opened this issue Dec 29, 2019 · 8 comments
Closed

长文本的效果忒儿差劲?这个怎么解决? #47

liuzhuang1024 opened this issue Dec 29, 2019 · 8 comments

Comments

@liuzhuang1024
Copy link

No description provided.

@liuzhuang1024
Copy link
Author

image

@zhang0jhon
Copy link
Owner

竞赛数据集没有相似的文档长文本,需要用新的数据集重新训练。

@Wo-Ki
Copy link

Wo-Ki commented Apr 2, 2020

我拿公开数据集(art+rects)训练430000次,用来识别车牌,效果也是不太好。前两个字还行,后面完全识别不到。这也是长文本的问题?

结果如下:
['皖', 'K'] [0.7606085 0.9683497 0.95730805]
['皖', 'K'] [0.7606051 0.9683497 0.9573083]
['啵', 'B'] [0.22663411 0.96232164 0.61064464]
['皖', 'G', '▪'] [0.4462814 0.99723685 0.8549502 0.9544695 ]
['皖', 'G', '▪'] [0.4462814 0.99723685 0.8549502 0.9544695 ]
['皖', 'A'] [0.8662637 0.84521544 0.9425101 ]
['啵', 'B'] [0.22663411 0.96232164 0.61064464]
['皖', 'A'] [0.8662637 0.84521544 0.9425101 ]

@zhang0jhon
Copy link
Owner

我把新的检测模型放到demo.sceneinception.com,可以先尝试一下效果

@Wo-Ki
Copy link

Wo-Ki commented Apr 2, 2020 via email

@heromanba
Copy link

我把新的检测模型放到demo.sceneinception.com,可以先尝试一下效果

您好,感谢您分享这么好的工作,请问有可以finetune的权重吗?docker里的pb好像是不可以训练的?

@Wo-Ki
Copy link

Wo-Ki commented Apr 3, 2020

能分享demo.sceneinception.com里的可以finetune的模型吗?非常感谢

@Nick198903
Copy link

我拿公开数据集(art+rects)训练430000次,用来识别车牌,效果也是不太好。前两个字还行,后面完全识别不到。这也是长文本的问题?

结果如下: ['皖', 'K'] [0.7606085 0.9683497 0.95730805] ['皖', 'K'] [0.7606051 0.9683497 0.9573083] ['啵', 'B'] [0.22663411 0.96232164 0.61064464] ['皖', 'G', '▪'] [0.4462814 0.99723685 0.8549502 0.9544695 ] ['皖', 'G', '▪'] [0.4462814 0.99723685 0.8549502 0.9544695 ] ['皖', 'A'] [0.8662637 0.84521544 0.9425101 ] ['啵', 'B'] [0.22663411 0.96232164 0.61064464] ['皖', 'A'] [0.8662637 0.84521544 0.9425101 ]

你好 问下 这个概率是怎么回事呢 两个字符为什么有三个概率呢

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

5 participants