We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? # to your account
loss回传时突然从tensor转成地址了
The text was updated successfully, but these errors were encountered:
@fenghuoxiguozu sorry,才看到,一直没有登录。 这个仓库代码是没有经过测试的,所以可能会有很多问题。如果你基于voc训练,建议你从github上搜1个在voc上训练的star比较高,效果比较好的模型;或者去paper with code上查看在voc数据集上表现优异的模型及其代码链接。 我这里主要是想实现数据增强,但是由于工作原因,未完成测试,仅提供数据增强的一些idea。
Sorry, something went wrong.
另外voc里面有255这种边界框,每个区域的边界都是这种,有专门处理voc的,就是根据[N,C,H,W]把边界框过滤掉在loss计算时变成[N,X]的形式计算,我这里没有考虑他的这种情况。
No branches or pull requests
loss回传时突然从tensor转成地址了
The text was updated successfully, but these errors were encountered: