Open HaiLongEmb opened 5 years ago
i has a problen together, do someone have solutions?
This is not a problem at all. It was mentioned in the paper that the class output is a logistics regression, not a classification, to handle the case where an object may belong to two or more classes. For example, if the training data has human face, woman face, man face classes, the face may belong to two classes.
On Sun, Oct 6, 2019, 2:01 AM NgTuong notifications@github.com wrote:
i has a problen together, do someone have solutions?
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/qqwweee/keras-yolo3/issues/461?email_source=notifications&email_token=AFSNTUKJXCWOFERA35OTMDLQNF5L5A5CNFSM4HN54ERKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEAOCHGY#issuecomment-538715035, or mute the thread https://github.com/notifications/unsubscribe-auth/AFSNTUKXPXTZDZNVKBNGFNTQNF5L5ANCNFSM4HN54ERA .
@datduonguva, thank you but i have a defference problem is two overlapping box belong to class together
Then simply lowering IOU threshold on yolo.py, which will be used for non_max_suppression, would help.
On Sun, Oct 6, 2019, 8:14 AM NgTuong notifications@github.com wrote:
@datduonguva https://github.com/datduonguva, thank you but i have a defference problem is two overlapping box belong to class together
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/qqwweee/keras-yolo3/issues/461?email_source=notifications&email_token=AFSNTUL6HBEHEDHYFP7UGY3QNHJDJA5CNFSM4HN54ERKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEAOIUCY#issuecomment-538741259, or mute the thread https://github.com/notifications/unsubscribe-auth/AFSNTUNG565MQ5LIYJKWJSLQNHJDJANCNFSM4HN54ERA .
@datduonguva , thank you, it worked.
The problem is as follows:
Overlapping of two types of detection,On this acute problem, how to solve.
Many Thanks in advance.