About creator_tool.py

[4fee8fea]

trafficstars

Thanks you for remarkable code sharing!

I have encountered a question here:

In creator_tool.py line#94:

roi = np.concatenate((roi, bbox), axis=0)

Why concatenate here? The latter functions calculate the iou matrix between roi and bbox,then the result is used to count the positive samples' total numbers.As if bbox has been to calculate iou matrix with itself.

Thank you for answering my question!

[Vickeyhw]

hi~ I'm confused by the same issue with you ,have you figure out?

[4fee8fea]

hh, I have not sloved the question,but I think the error seems to be an obvious logical error. Maybe we can certify it by delete the line and train again. What confuses me is that by adding the line, author gets good mAP. Maybe by deleting the line,we can get the same performance faster.

Vickeyhw [email protected] 于2019年8月10日周六 上午10:54写道：

hi~ I'm confused by the same issue with you ,have you figure out?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/chenyuntc/simple-faster-rcnn-pytorch/issues/143?email_source=notifications&email_token=AMYSJ33ZWR5E77K4ARWKCBTQDYUVZA5CNFSM4IKG5PU2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD4AEPPI#issuecomment-520112061, or mute the thread https://github.com/notifications/unsubscribe-auth/AMYSJ3YQ5XKNNHVMT7JNPFDQDYUVZANCNFSM4IKG5PUQ .

[sgq1001]

I also have the same question about this, it seems useless and logically false