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

Did you try to cascaded with OHEM or just use OHEM in the third stage? #32

Closed
HiKapok opened this issue Jul 6, 2018 · 6 comments
Closed

Comments

@HiKapok
Copy link

HiKapok commented Jul 6, 2018

OHEM is useful in my experiment, did you try to combine OHEM with cascaded rcnn? should I apply OHEM in every stage of rcnn?

@zhaoweicai
Copy link
Owner

I didn't try OHEM. It is orthogonal. Let me know if you find OHEM works in cascade rcnn.

@HiKapok
Copy link
Author

HiKapok commented Jul 7, 2018

@zhaoweicai ok, thanks for your reply

@HiKapok HiKapok closed this as completed Jul 7, 2018
@yaoqi-zd
Copy link

yaoqi-zd commented Aug 1, 2018

@HiKapok Hi, did you try OHEM with cascade rcnn, did OHEM bring any performance gain?

@HiKapok
Copy link
Author

HiKapok commented Aug 1, 2018

@yaoqi-zd no, it's worser with OHEM, I have tried to apply OHEM in the third stage or in all three stage, the former is better, but still worser than the one without OHEM

@yaoqi-zd
Copy link

yaoqi-zd commented Aug 1, 2018

@HiKapok
I have tried to apply OHEM only in the first stage,
dataset: kitti 2d detection
classes: car, pedestrian and cyclist
ohem param:
rois_per_img: 256
performance is sligted better on easy samples but the same on moderate and hard samples.

@HiKapok
Copy link
Author

HiKapok commented Aug 1, 2018

@yaoqi-zd thank you for your information. I did not get any positive results on my dataset when OHEM is applied.

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

3 participants