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

关于对比方法 #4

Open
wzhwantmoney opened this issue Apr 1, 2024 · 2 comments
Open

关于对比方法 #4

wzhwantmoney opened this issue Apr 1, 2024 · 2 comments

Comments

@wzhwantmoney
Copy link

作者您好,请问您DETR和Deformable-DETR模型在BCCD上数据集的数据是自己实验得到的吗?我使用DINO模型在BCCD数据集上的精度仅达到了59.5,请问您是怎样训练的?按照这几个模型原论文中的精度来看,DINO要远优于DETR和Deformable-DETR,所以我很疑惑为什么在BCCD数据集上,他们的性能有这么大差距。并且哪怕使用了swin-Transformer作为backbone,也达不到您论文中两个方法的精度。我也使用了在coco数据集上预训练的权重,我的BCCD数据集也同样划分为8:2(72张验证集)。不胜感激。以下是我的实验结果。
DINO-resnet

@wzhwantmoney
Copy link
Author

我用Deformable-DETR在BCCD数据集上进行了实验,训练了100个epoch,最后的mAP仅仅有48左右,请问您论文中63.5的精度是如何得到的?还有您方便提供一下相关实验的权重文件吗?

@xqq777
Copy link

xqq777 commented Jul 12, 2024

我在BCCD数据集上进行了实验,训练了100个epoch,最后的mAP只有48左右,请问您论文中63.5的精度流量获得吗?还有您方便提供一下相关实验的权重文件吗?

臣附议

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