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

the parameters for OIS/ODS #10

Open
ShiAngWang opened this issue Dec 9, 2021 · 7 comments
Open

the parameters for OIS/ODS #10

ShiAngWang opened this issue Dec 9, 2021 · 7 comments

Comments

@ShiAngWang
Copy link

Hi, in your paper, you mentioned "Optimal dataset scale (ODS) usually refers to the best performance when selecting optimal parameters for the whole dataset, while optimal image scale (OIS) refers to the best performance when selecting special parameters for each image. "
Are the parameters here refering to the merging threshold or the network parameters?

@ShiAngWang
Copy link
Author

ShiAngWang commented Dec 9, 2021

Like in the gPb, they output a edge probability map, so the threshold for determing whether a pixel is in edge is the parameter would be tuned in generating Pb, Rb and Fb. So in this context, ODS refers to the best performance when selecting optimal the threshold for the whole dataset, and OIS refers to the best performance when selecting the threshold for each image.

But in your method, the network should output a deterministic partition for a graph in a specific superpixel merging threshold. So the parameters tuned for ODS and OIS is the superpixel merging threshold?

@ShiAngWang
Copy link
Author

ShiAngWang commented Dec 9, 2021

您好,十分感谢您之前对我的回复!
我的这个问题是说,因为 del 这个任务本身是一个分割任务,而不是一个输出 边缘概率图 的边缘检测任务。对于边缘检测任务的 ODS 或 OIS 来说,需要调节的参数就是边缘概率的阈值,进而得到边缘检测的精确率、召回率、P-R 曲线、F measure 等。
而对于分割来说,本身输出的就是一个特定的分割区域,不涉及到边缘阈值这个概念,所以在 del 中,在计算 ods/ois 下的 Fb 时,调节的是什么参数呢,是指的是整个网络结构参数还是超像素融合的 threshold 呢?
以及我还比较疑惑的就是,因为您也评测了 slic 和 EGB 等超像素方法,它们本身也是分割方法而不是边缘检测方法,针对它们又是调节的什么参数,进而调整的精确率与召回率呢?

@ShiAngWang
Copy link
Author

简单来说,就是我对评价分割方法的 Fb 有些疑惑,不知道是调节哪个参数产生的 PR 曲线,是针对不同的方法有不同的参数吗?

以及想问下您, SEISM 中评价分割方法的代码是这个吗? https://github.com/jponttuset/seism/blob/master/src/measures/eval_segm.m

@ShiAngWang
Copy link
Author

我这边阅读了 “Learning to Detect Natural Image Boundaries Using Local Brightness, Color,and Texture Cues” 和 “Contour Detection and Hierarchical Image Segmentation”两篇文章,对 P-R curve for boundary 这个指标还是有上面的疑惑
就比如针对您这篇 del,调节是哪个参数以调节的精确率与召回率呢?

@yun-liu
Copy link
Owner

yun-liu commented Dec 12, 2021

调节superpixel merging的参数。除此之外,DEL也没其他参数了啊

@ShiAngWang
Copy link
Author

调节superpixel merging的参数。除此之外,DEL也没其他参数了啊

您好,我对这个 ods、ois 的评估过程还是有些疑惑,我想再请问下,在您的del这个工作中,对 ods 和 ois 这两种情况具体是怎么调节的参数呢,
ods 指的是全局最优,是通过不断调整全局统一的一个合并阈值,以观察并得到一个最高的 Fb、Fop 指标吗?
ois 指的是单张最优,这个是怎么调节的呢,是在每一张上尝试不同的阈值,达到一个单张最优的 Fb、Fop,然后再一张张试,为每一张设定一个不同的合并阈值吗?

@yun-liu
Copy link
Owner

yun-liu commented Dec 19, 2021

请您阅读一下这个代码:https://github.com/yun-liu/DEL/blob/master/examples/del/test_bsds_seg.ipynb

运行这个脚本,会生成不同参数下的分割结果,最后用SEISM的代码评测。

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