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

关于两篇论文中的结果存在出入的疑惑 #13

Open
youngbaldy opened this issue Jan 6, 2024 · 3 comments
Open

关于两篇论文中的结果存在出入的疑惑 #13

youngbaldy opened this issue Jan 6, 2024 · 3 comments

Comments

@youngbaldy
Copy link

youngbaldy commented Jan 6, 2024

首先感谢您的工作!我在阅读SAMRS与RSP(An Empirical Study of Remote Sensing Pretraining)时,发现两篇文章公布的rsp-r50分割结果有所差异。
具体而言:
(1) SAMRS中表3 rsp-r50在potsdam结果为 OA=90.49 mF1=90.97
QJHWYP@5@TO_M8B4J7M60SG
RSP中表6 rsp-r50的结果为OA=90.61 mF1=89.94
HT%L8E}CYS7S首先感谢您的工作!我在阅读SAMRS与RSP(An Empirical Study of Remote Sensing Pretraining)时,发现两篇文章公布的rsp-r50分割结果有所差异。 具体而言: (1) SAMRS中表3 rsp-r50在potsdam结果为 OA=90.49 mF1=90.97 QJHWYP@5@TO_M8B4J7M60SG RSP中表6 rsp-r50的结果为OA=90.61 mF1=89.94 B)SKSJI}0
(2) SAMRS中表4 rsp-r50在isaid结果为mIoU为32.97
Z~W$N1A_WI{JR9UZ NXI@7B
RSP中表7 rsp-r50的结果为mIoU为61.6
GIFU$R_T1VHR_CY)TZS6 HB
我不太清楚造成这种差异的原因是什么?特别是在isaid上的结果差异比较大,我有点不太确定该以哪篇文章的结果为准。期待您的回复。

@DotWang
Copy link
Collaborator

DotWang commented Jan 6, 2024

@youngbaldy
这个原因是这样:

  1. Potsdam在RSP那篇文章中用的是NIR-R-G,后来我按照CV那边的原则都换成RGB了,所以重跑了一下,这个差距不大
  2. 之前RSP那篇文章统一用的老版本mmsegmentation跑的,后来我环境换了,而且openmmlab系列大改,现在都搞成新版本了,但在用新版本之前,32.97这个好像是用我目前这套代码跑出来的,optimizer,scheduler什么的都跟之前不一样,我觉得你还是以mmseg的为准吧,isaid数据集还是比较敏感的

@youngbaldy
Copy link
Author

感谢您的回复!您是指以老版本mmseg的RSP codebase为准吗?

@DotWang
Copy link
Collaborator

DotWang commented Jan 7, 2024

@youngbaldy

是的,RSP的benchmark以RSP原论文和对应repo为准

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