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

请问一下Dense teacher提出的DPL是否其实只是针对预测的类别分支? #39

Open
joeyslv opened this issue May 30, 2023 · 2 comments

Comments

@joeyslv
Copy link

joeyslv commented May 30, 2023

在图1中,有Dense Pseudo-label 和 Pseudo-box Label两个分支,分别计算了Ldpl损失和Lbox损失,所以Ldpl其实就是对应教师模型输出分类分支后的sigmoid所做的吗?整个模型还是用了NMS来生成伪框监督回归分支的损失计算?
3.3 Dense Pseudo-Label
Since the learning region is selected, unsupervised learning for regression branch can be easily achieved.
3.3中关于回归的分支的叙述较少,上面说由于选择了学习区域,因此回归分支的学习更容易实现,这个更容易实现指的是什么呢?
希望作者能够解答一下
感谢

@ZRandomize
Copy link
Collaborator

图一只是一个我们方法和已有方法的比较,我们只使用了Ldpl这个loss
回归更容易实现是因为如果没有区域选择步骤,背景区域的teacher reg是没有意义的,使用这些信息学习会导致性能下降。而区域选择可以有效的筛选出reg有意义的部位

@joeyslv
Copy link
Author

joeyslv commented Jun 6, 2023

图一只是一个我们方法和已有方法的比较,我们只使用了Ldpl这个loss 回归更容易实现是因为如果没有区域选择步骤,背景区域的teacher reg是没有意义的,使用这些信息学习会导致性能下降。而区域选择可以有效的筛选出reg有意义的部位

好的,大概了解了,感谢

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

2 participants