search

houqb / DSS

code for "Deeply supervised salient object detection with short connections" published in CVPR 2017
MIT License
239 stars 79 forks source link

the test performance is inconsistent with the result in paper #2

Closed ArcherFMY closed 7 years ago

ArcherFMY commented 7 years ago

Hi, Andrew,

Thank you very much to provide the code to the paper DSS.

I followed the instructions to download the code and the pretrained model. Then I used the provided code DSS-tutorial.ipynb to run test on ECSSD dataset. The PR curve I got is even worse than DCL.

Any idea about what happened? Thank you in advance.

houqb commented 7 years ago

Not sure what's going on in your case. Actually the released model is not my initial one, but the results are very close to ones we reported in our paper. Could you show me your results (including Fm and MAE)?

ArcherFMY commented 7 years ago

plz see here. https://github.com/ArcherFMY/my-dss-result

the threshold is 2*mean(saliency map)

houqb commented 7 years ago

It seems the results you provided are different from mine. 0001

Maybe this is due to different versions of Caffe.

ArcherFMY commented 7 years ago

I used the newest Caffe version.

Would you mind sharing your caffe? Then I can test again.

houqb commented 7 years ago

I have released. You can find it in my another repo.

Kind regards, Qibin (Andrew) Homepage: http://mmcheng.net/qbhou/

On Thu, Jul 20, 2017 at 2:01 PM, ArcherFMY notifications@github.com wrote:

I used the newest Caffe version.

Would you mind sharing your caffe? Then I can test again.

— You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub https://github.com/Andrew-Qibin/DSS/issues/2#issuecomment-316605333, or mute the thread https://github.com/notifications/unsubscribe-auth/ANom8xxqUo3Hht8GD6xI50Aw2srtxEKrks5sPu1VgaJpZM4OdlZy .

ArcherFMY commented 7 years ago

here is my results... selection_003

houqb commented 7 years ago

good

houqb commented 7 years ago

If you average the results from so2, so3, so4, and fusion, the scores would be better.

Kind regards, Qibin (Andrew) Homepage: http://mmcheng.net/qbhou/

On Thu, Jul 20, 2017 at 2:22 PM, ArcherFMY notifications@github.com wrote:

here is my results... [image: selection_003] https://user-images.githubusercontent.com/11482534/28403173-1b168a60-6d56-11e7-9224-507edc44806b.png

— You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub https://github.com/Andrew-Qibin/DSS/issues/2#issuecomment-316608668, or mute the thread https://github.com/notifications/unsubscribe-auth/ANom8yvS50vF6yW1aQyqE8pHaGWrpYqEks5sPvJBgaJpZM4OdlZy .